# aruco pose estimation on pi 4 with ROS melodic
platform: Canonical official ubuntu-server 18.04 built for armhf/arm64 pi4

:::info
ubuntu 20.04 failed.
because ubuntu 20.04 is for ROS noetic, and noetic move to python3 (python 2 EOL at 1/1/2020), but cv_bridge uses libpython.boost which don't support python3
:::

:::info
ubuntu mate don't have binary image built of 18.04 for pi4
only canonical ubuntu-server has 18.04 built for pi4, but it's server version, so have to install desktop environment manually
:::

---

https://cdimage.ubuntu.com/releases/18.04/release/
ubuntu-18.04.5-preinstalled-server-arm64+raspi4.img.xz

---

sudo apt install ubuntu-mate-desktop

---

ROS apt source & key
http://wiki.ros.org/melodic/Installation/Ubuntu

sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'
sudo apt-key adv --keyserver 'hkp://keyserver.ubuntu.com:80' --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654

sudo apt update

---

sudo apt install ros-melodic-desktop-full

---

source /opt/ros/melodic/setup.bash

(add at the end of ~/.bashrc)

---

sudo apt install python-catkin-tools
\# for `catkin` command

---

all done!

rospy & cv_bridge is installed by default!
(remember to source ros setup.bash file)

---

sudo apt install ros-melodic-usb-cam

---

#!/usr/bin/env python

import rospy
import sensor_msgs
from cv_bridge import CvBridge, CvBridgeError
import cv2
import cv2.aruco
import numpy as np

rospy.init_node('aruco_pose', anonymous=True)

tag_dict = cv2.aruco.getPredefinedDictionary(cv2.aruco.DICT_4X4_50)

image_pub = rospy.Publisher('image_out', sensor_msgs.msg.Image)
bridge = CvBridge()
def callback(data):
    img = bridge.imgmsg_to_cv2(data, 'bgr8')
    #cv2.imshow('raw', img)
    #cv2.waitKey(1)
    (corners, tag_ids, rej_corners) = cv2.aruco.detectMarkers(img, tag_dict)
    #print(corners)
    #print(tag_ids, reject_tags)
    #cv2.aruco.drawDetectedMarkers(img, rej_corners, np.array([]), np.array([255, 255, 0]))
    cv2.aruco.drawDetectedMarkers(img, corners, tag_ids, np.array([0, 255, 0]))
    image_pub.publish(bridge.cv2_to_imgmsg(img, 'bgr8'))

image_sub = rospy.Subscriber("/usb_cam/image_raw", sensor_msgs.msg.Image, callback)
rospy.spin()

---

full code:
https://github.com/wolfdigit/aruco_pose

---

sudo apt install git

mkdir -p catkin_ws/src
cd catkin_ws/src
git clone https://github.com/wolfdigit/aruco_pose.git

cd ..
catkin build

source devel/setup.bash

---

tag generator: https://chev.me/arucogen/

---

opencv imshow & waitkey is laggy!
use ros imgmsg publish /w image_view instead