Spool Autodetection using Apriltags Proof-of-Concept

[marcocarnut]

I often forget to select the correct spool, messing all measurements. I initially thought of putting a new camera near the spool to read the QR code I glued on it, but QR code reading is somewhat unreliable -- rather minor twists on the spool or unfavorable lighting conditions would upset the QR code detection.

Apriltags, however, are very robust -- they're widely used in robotics, being specifically designed to be reliably detected at a distance, at weird angles and unfavorable lighting. They can only carry a small numeric ID, but that's more than enough.

So I printed a ring of Apriltags, glued on the spool, took a picture with the same camera I use to monitor the print, getting something like that:

Then I ran the image through the Apriltag "demo" utility, and, to my joy, it detected the tags perfectly, spitting out its ID (which I made the same as my spool databaseId).

Not wanting to mess with the plugin's source code just yet, I wrote the small bash script below to detect and automatically select the spool via the API:

#!/bin/bash

# Create a self-lock so we don't execute more than once

[ "${FLOCKER}" != "$0" ] && exec env FLOCKER="$0" flock -en "$0" "$0" "$@" || :

tmpdir=/dev/shm
imgfile=$tmpdir/snap.jpg
url='http://localhost:8080/?action=snapshot'
action_url=http://localhost/plugin/SpoolManager/selectSpool
api_key=####put-your-api-key-here#####
attempt=0
maxattempts=80
interval=2
echo "Autoselect starting at $(date)"

while [ "$attempt" -ne "$maxattempts" ] ; do
  attempt=$(( attempt + 1 ))
  echo "Attempt $attempt of $maxattempts... " | tee -a  $tmpdir/autoselect.log
  # Get a snapshot from the camera
  wget -O $tmpdir/snap-$attempt.jpg "$url" 2>/dev/null
  # Try to identify it
  spool_ids=( $(
    /home/pi/apriltag_demo -f tagCustom48h12 $tmpdir/snap-$attempt.jpg 2>/dev/null | \
      perl -ne 'print "$1\n" if m{detection.*-(\d+)}o' | \
      sort -u
  ) )

  if [ "${spool_ids[1]}" != "" ] ; then
     echo 'Ambiguous detection, this should not happen!' | tee -a  $tmpdir/autoselect.log
     exit 3
  fi
  echo spool_ids=$spool_ids
  spool_id=${spool_ids[0]}
  if [ "$spool_id" == "" ] ; then
    echo 'Spool identifier tag not detected' | tee -a  $tmpdir/autoselect.log
    sleep $interval
    continue
  fi

  echo "Found identifier: $spool_id" | tee -a  $tmpdir/autoselect.log

  data="{\"databaseId\":$spool_id,\"toolIndex\":0,\"commitCurrentSpoolValues\":false}"
  json="Content-Type: application/json"
  sid=$( curl -X PUT -d "$data" -H "$json" "$action_url?api_key=$api_key" 2>/dev/null | jq -r .selectedSpool.databaseId )
  if [ "$sid" -eq "$spool_id" ] ; then
    echo 'Success!' | tee -a  $tmpdir/autoselect.log
    exit 0
  else
    echo "Failed when trying to set spool in Octoprint" | tee -a  $tmpdir/autoselect.log
    exit 1
  fi
done

echo "Exceeded maximum number of attempts" | tee -a  $tmpdir/autoselect.log
exit 2

At least in my setup, it works just fine and I no longer have to worry about manually selecting the correct spool :) Now my usage measurements are always correct (at least with the new spools after I started using this).

It would be nice to have this feature properly coded in the plugin natively. I'm a C, perl and nodejs guy, but I'm a relative noob in Python and a total noob in Octoprint plugin development, so I'd appreciate any guidance/suggestions you might have.

If anyone wants to reproduce this, you'll need wget, curl, jq and apriltag_demo from the examples directory of:

• https://github.com/AprilRobotics/apriltag

[github-actions[bot]]

This issue has been automatically marked for closing, because it has not had activity in 30 days. It will be closed if no further activity occurs in 10 days.

[github-actions[bot]]

This issue was closed, because it has been already marked for 10 days with no activity.