Introduction to AR Tags

Requirements

This practical requires the library opencv-contrib-python, which has additional modules that are not included in opencv-python. Install with the following command:

pip3 install opencv-contrib-python

Generating AR Tags

You can generate and print your own AR tags for any use case. ArUco contains a number of AR Tag Libraires:

• DICT_4X4_100
• DICT_4X4_1000
• DICT_4X4_250
• DICT_4X4_50
• DICT_5X5_100
• DICT_5X5_1000
• DICT_5X5_250
• DICT_5X5_50
• DICT_6X6_100
• DICT_6X6_1000
• DICT_6X6_250
• DICT_6X6_50
• DICT_7X7_100
• DICT_7X7_1000
• DICT_7X7_250
• DICT_7X7_50
• DICT_APRILTAG_16H5
• DICT_APRILTAG_16h5
• DICT_APRILTAG_25H9
• DICT_APRILTAG_25h9
• DICT_APRILTAG_36H10
• DICT_APRILTAG_36H11
• DICT_APRILTAG_36h10
• DICT_APRILTAG_36h11
• DICT_ARUCO_ORIGINAL

The tags of format DICT_{SIZE}x{SIZE}_COUNT are ArUco based tags that use SIZExSIZE bits for tag information and have COUNT distinct tags. We will be using the ArUco original librariy to create two tags:

import numpy as np
import cv2
arucoDict = cv2.aruco.Dictionary_get(cv2.aruco.DICT_ARUCO_ORIGINAL)
SIZE = 500 # pixels
marker = np.zeros((SIZE, SIZE, 1), dtype=np.uint8)
ID = 0
cv2.aruco.drawMarker(arucoDict, ID, SIZE, marker, 1)
cv2.imwrite('DICT_ARUCO_ORIGINAL_id_{}_{}.png'.format(ID, SIZE), marker)

CHALLENGE 1: Generating AR Tags

Generate AR tags for the dictionary DICT_APRILTAG_16H5 for ids 7, 18, and 23.

• These should be of size 500 pixels by 500 pixels.

Detecting AR Tags

The ArUco library has built-in functionality for detecting AR tags within images:

tags = cv2.imread('data/two_tags_ARUCO_ORIGINAL.png')
arucoDict = cv2.aruco.Dictionary_get(cv2.aruco.DICT_ARUCO_ORIGINAL)
corners, ids, rejects = cv2.aruco.detectMarkers(cv2.cvtColor(tags, cv2.COLOR_BGR2GRAY), arucoDict)
detection = cv2.aruco.drawDetectedMarkers(tags, corners, borderColor=(255, 0, 0))
cv2.imwrite('detection_two_tags_ARUCO_ORIGINAL.png', detection)

CHALLENGE 2: Calculating Distance Between Tags

Calculate the distance between the AR tags in the file data/two_tags_APRILTAG_16H5.png in centimeters. Some notes to keep in mind:

• The real-world width of each tag is 3.
• The dictionary being used is APRILTAG_16H5.
• The image view is exactly perpendicular to the camera.
• Both cv2 and numpy are useful packages here.