List of ArUco and ChArUco IDs

This page documents the ID's and sizes of ArUco and ChArUco boards used by Stretch 4.

ArUco Tags

ID	Outer Marker Size	Inner Marker Size	Dictionary	Used By
29	30mm	20mm	DICT_6x6_100
30	50mm	40mm	DICT_6x6_100
31	80mm	60mm	DICT_6x6_100
32	100mm	80mm	DICT_6x6_100
33	120mm	100mm	DICT_6x6_100
43	50mm	40mm	DICT_6X6_50	Docking Station, Right Tag
44	50mm	40mm	DICT_6X6_50	Docking Station, Left Tag
45	50mm	40mm	DICT_6X6_50	Docking Station, Apex Tag
66	32mm	25mm	DICT_6x6_100	Wrist Yaw Joint
200		14mm	DICT_6x6_250	Left Fingertip
201		14mm	DICT_6x6_250	Right Fingertip

ChArUco Boards

Board Type	Board Size	# Rows	# Columns	Checker Size	Marker Size	Dictionary	Start ID	End ID	Used By
White ArUcos	280x216mm	5	7	37mm	27mm	DICT_4X4_50	0	17	Camera Calibration

*Board type is whether the board is inscribed in white or black squares.

How to read these tables

For ArUco tags, outer and inner marker size refers to the outer white boundary and inner black boundary of the tag. A 50mm outer with a 40mm inner means the tag has a 5mm white border around the tag. For ChArUco boards, checker size refers to square length of each black/white square in the grid. Marker size refers to the inner black boundary of the tag. Hence, a 37mm checker size with a 27mm marker size means the tags have a 5mm border around the tag. ChArUco boards come in two variants; the ArUco marker can be inside the black or white squares. For example, the camera calibration ChArUco board uses the white variant.

ArUco detection requires a "dictionary". DICT_6X6_50 refers to a ArUco tag with 6x6 grid that is capable of representing 2³⁶ IDs. The _50 refers to the number of IDs the dictionary can actually detect (0-49 in this case). Smaller dictionaries have larger Hamming distance between each ID, and therefore better error correction and reliability, allowing for better detection on blurry or slightly occluded images of the tag. The dictionary options are:

• cv2.aruco.DICT_nXn_50 (Contains IDs 0-49)
• cv2.aruco.DICT_nXn_100 (Contains IDs 0-99)
• cv2.aruco.DICT_nXn_250 (Contains IDs 0-249)
• cv2.aruco.DICT_nXn_1000 (Contains IDs 0-999)

Larger dictionaries are supersets of smaller dictionaries. This means tag ID 0 will be visible in all of these dictionaries.

You might expect larger NxNs to increase Hamming distance, but it's actually the opposite. A 4x4 marker only has 16 internal bits. The absolute maximum Hamming distance possible between any two markers is 16 (and practically, it's much lower). A 7x7 marker has 49 internal bits. This gives the dictionary more combinations to pick from, allowing it to select 50 markers that are radically different from one another, resulting in a much higher Hamming distance and better error correction capability. OpenCV supports

• 4x4 (cv2.aruco.DICT_4X4_...)
• 5x5 (cv2.aruco.DICT_5X5_...)
• 6x6 (cv2.aruco.DICT_6X6_...)
• 7x7 (cv2.aruco.DICT_7X7_...)

The tradeoff is fitting more bits into the same physical size marker requires the camera to have high enough resolution to see each bit clearly enough. For example, a 1280p camera (or out-of-focus camera) might struggle to see a 7x7 marker 2 meters away, but not with a 4x4 marker.

OpenCV's cv2.aruco module also support AprilTags, which was developed at UMich and is said to have better robustness to lighting/angles. The dictionaries are listed below, although, we do not use them for Stretch 4.

• DICT_APRILTAG_16h5 (Effectively 4x4)
• DICT_APRILTAG_25h9 (Effectively 5x5)
• DICT_APRILTAG_36h10 / 36h11 (Effectively 6x6)