Skip to content

Coded circular targets: 12-bit markers, printing, sizing, and the 14-bit / 16-bit / 20-bit family

  • Status: unverified
  • Applies to: Metashape Pro 2.x — and PhotoScan / Metashape 1.x via the same CircularTarget* enum on the legacy Metashape.TargetType (and PhotoScan.TargetType). AprilTag detection is Metashape Pro 2.2+; see the AprilTag article.
  • Edition: Pro
  • Diátaxis: how-to
  • Confidence: medium
  • Last reviewed: 2026-06-05

Confidence: medium. The four-family CircularTarget bit-count enumeration, the printing workflow, and the "12-bit decoded more precisely" claim are sourced from the official Metashape Pro 2.3 User Manual (p. 106–107). The sizing rules are forum-attested via a community user's empirical study; the 30-px upper limit is sourced from Agisoft's own Coded Targets & Scale Bars in PhotoScan Pro 1.0.0 tutorial. The detection-angle behavior is partially attested but warrants further empirical work.

The marker-separation, masquerade, and best-subset figures added below come from analyzing the decoded CircularTarget families with a rotation-invariant Hamming distance; worst-pair values are exact for the 12/14/16-bit families and greedy lower bounds for 20-bit.

Problem

You need to place printed targets in a scene for one of:

  • Coordinate-system anchors — coded targets give the bundle automatic, label-stable correspondences for georeferencing.
  • Scale anchors — two coded targets at known distance define a scalebar.
  • Manual-tie-point fall-back — coded targets are 100% valid matches that supplement automatic feature matching (see Helping alignment when photos don't align).

Coded circular targets — the family Metashape has supported since PhotoScan 1.0 — are still the default choice for close-range photogrammetry: cheap to print, sub-pixel detection accuracy, automatic ID assignment.

This article covers the 12-bit / 14-bit / 16-bit / 20-bit circular family specifically. For the AprilTag family added in 2.2.0, see AprilTag detection — choosing a variant.

What 12-bit circular coded targets are

The manual:

"The following types of coded targets are detected in Metashape: Circular and AprilTag. […] Metashape supports four types of circle CTs: 12 bit, 14 bit, 16 bit and 20 bit. While 12 bit pattern is considered to be decoded more precisely, 14 bit, 16 bit and 20 bit patterns allow for a greater number of CTs to be used within the same project." — Metashape Pro User Manual 2.3, p. 107

A 12-bit circular target is a small printed pattern with:

  • A solid black central circle (the centroid is the detected 2D point).
  • A surrounding ring divided into black/white segments encoding 12 bits of identifier — yielding 161 unique codes after parity-check filtering in the current Print Markers output (561 for 14-bit, 2,001 for 16-bit, and 26,013 for 20-bit).

The detection algorithm finds the central circle by ellipse fitting (so the same target works at varying viewing angles), then decodes the ring to recover the integer ID.

Origin and patent

The 12-bit circular coded-target design is based on a method described in:

Schneider, C. T. (1991). "3-D Vermessung von Oberflächen und Bauteilen durch Photogrammetrie und Bildverarbeitung." Proc. IDENT/VISION 91, 14–17.

The corresponding German patent has expired, so the design is freely usable. Independent open-source implementations of related coded targets (including 14-bit) exist:

The four bit-count variants

Variant API enum (Metashape.TargetType) Library size Suggested use
12-bit circular CircularTarget12bit 161 codes Default close-range; "most precise" per manual
14-bit circular CircularTarget14bit 561 codes Larger projects than 12-bit can name
16-bit circular CircularTarget16bit 2,001 codes Industrial / many-target metrology
20-bit circular CircularTarget20bit 26,013 codes Massive projects (rare)

The exact code counts vary by Metashape version and the parity-check scheme in effect; the manual's own description ("12 bit pattern is considered to be decoded more precisely, 14 bit, 16 bit and 20 bit patterns allow for a greater number of CTs") implies the families differ by order of magnitude in library size, which the table reflects.

To check the exact count for any variant on your installed Metashape version, open Tools → Markers → Print Markers…, select the variant, and count the targets in the generated PDF. (For close-range projects with fewer than ~25 unique markers, 12-bit is the default choice; for anything larger, verify before printing.)

The manual's claim that 12-bit is decoded "more precisely" relates to the lower bit-count's better signal-to-noise ratio in the ring-decoding step — the parity-checked ring is shorter and survives more degradation.

Printing the targets in Metashape

The targets are generated by Metashape itself:

Demo verified: ✗ — pending Tier 3 reproduction on a real Metashape install.

  1. Open Metashape Pro (target generation is Pro-only).
  2. Tools → Markers → Print Markers… opens the Print Markers dialog.
  3. Pick the target type (12-bit, 14-bit, 16-bit, 20-bit, or AprilTag if on 2.2+) and choose a target size in millimeters.
  4. OK writes a multi-page PDF; the first page contains all the unique codes.
  5. Print the PDF on a paper / vinyl that won't reflect or warp; mount on rigid backing for field deployment.

The printed PDF is the canonical source — there's no need to generate targets externally unless you need a non-default layout (turntables, custom plates, embedded references). For turntables specifically, the community JSFiddle target generator produces a printable circular layout; one open issue: the default count = Math.trunc(circ / (codeSize * 3)) formula in the turntable() function over-spaces targets; replace with count = Math.trunc(circ / (codeSize + padding)) for evenly-spaced rings.

Optimal target size (forum-attested empirical guidance)

How big should the printed target be? The detection algorithm has a minimum size below which the central circle isn't reliably detected, and an upper limit above which it becomes wasteful (or reduces the count of usable targets in the field of view).

Lower bound — minimum 9–10 pixel central circle. A community user's empirical study with 12-bit targets on multiple cameras (DJI Phantom 4, GoPro Hero 4):

"In order to be detected, the center point of a PhotoScan 12-bit target has to cover a minimum of 9–10 pixels on the image. […] As the global diameter of the target is 3.5 times the diameter of the center point, the whole target should have a minimum size of 35 cm [for a 1 cm GSD]." — Yoann Courtois (community forum user), 2017-04-03, PhotoScan 1.3 (permalink)

Upper bound — the central circle should not exceed 30 pixels. From Agisoft's own Coded Targets & Scale Bars in PhotoScan Pro 1.0.0 tutorial, as quoted in the same forum thread:

"The central black circle-point on the taken photo is not greater than 30 pix." — Agisoft, Coded Targets & Scale Bars tutorial (PhotoScan Pro 1.0.0), via forum thread t=2768

So aim for 10–30 pixels of central circle on the image, with the global target diameter ≈ 3.5× the central-circle diameter.

Worked example — sizing for a known GSD

If your project's expected GSD is 1 cm:

Quantity Formula Value
Minimum central-circle diameter on image 10 px (forum-attested floor) 10 px
Minimum central-circle diameter at GSD 10 px × GSD 10 × 1 cm = 10 cm
Minimum global target diameter 3.5 × central diameter 3.5 × 10 cm = 35 cm
Recommended print size of central circle ~5 px × GSD margin above floor ~5 × 1 cm = 5 cm (use as a working radius to leave headroom; resulting central diameter on print is ≥ 10 cm)

If GSD is 2 mm (close-range artefact scanning), the central circle becomes 2 cm and the global target ~7 cm — easy to print on letter paper. If GSD is 5 cm (high-altitude aerial), the global target balloons to 175 cm, which is impractical to lay out — at which point the manual's recommendation becomes relevant:

"while [coded targets] generally prove to be useful in close-range imagery projects, aerial photography projects will demand too huge CTs to be placed on the ground, for the CTs to be detected correctly. Hence, for aerial photography projects it makes sense to consider non-coded targets implementation." — Metashape Pro User Manual 2.3, p. 107

Detection-angle behavior

Detection from oblique angles is partially understood:

"I didn't handle precise measurements but I can say the angle of view is not the main limitation in automatic detection. For UAV survey with nadiral images for example, the limit of detection will be the effective GSD on the target when the camera will fly away from the target position, and not the angle of view." — Yoann Courtois (community forum user), 2017-04-03, PhotoScan 1.3 (permalink)

In other words: the detector copes with oblique angles reasonably well (the central-circle ellipse fit is designed for it), but the effective GSD on the target's plane is what drives detection success. A target viewed at 60° projects onto the image as an ellipse with the minor axis foreshortened by cos(60°) ≈ 0.5 while the major axis is preserved. The ellipse's minor-axis length on image falls below the 10-px minimum sooner than a nadir target would; the ellipse-fit detector is robust to the foreshortening up to the point where the minor axis is too short to anchor the fit reliably.

This is an experimental area worth testing on your specific camera + flight plan. Print targets at multiple sizes, capture from multiple angles, measure the major and minor axes of the central-circle ellipse on each image, and note the detection threshold. The "10 px minor-axis floor" gives a starting point, but no upper-bound for θ has been forum-attested in the corpus.

Marker ambiguity and the minimum Hamming distance

The separation, masquerade, and best-subset figures in the sections below come from analyzing the decoded CircularTarget families — the patterns in the Print Markers PDFs — with a rotation-invariant Hamming distance. Worst-pair (max-min) values are exact for the 12-, 14-, and 16-bit families and greedy lower bounds for 20-bit.

Coded circular targets are not all equally distinguishable from one another. Whether two markers can be confused is governed by a rotation-invariant Hamming distance: the number of ring segments that must be misread to turn one valid code into the other, minimized over every rotation of the ring. A printed target can be photographed at any orientation, so the relevant distance is the smallest one across all N cyclic rotations:

d(a, b) = min over r in 0..N-1 of  hamming( code_a, rotate(code_b, r) )

The family floor is the smallest d(a, b) over all pairs in the library — the worst case for the whole family: how many segment errors are needed before some pair collides.

For the CircularTarget families the floor is 2. Agisoft's families apply no inter-marker distance optimization: the codes are simply the valid parity-checked patterns, and the closest pairs sit only two segments apart at their nearest relative rotation. Two consequences follow:

  • A single-segment (1-bit) error is always safe. No two distinct markers differ by only one segment, so one misread segment can never turn a valid marker into another valid marker — it produces an invalid pattern that the parity check rejects (the marker is dropped, not mis-identified).
  • A two-segment (2-bit) error can cause a silent mis-identification. For the closest pairs (those at the floor distance of 2), two coincident segment errors — from blur, low resolution, a printing defect, or partial occlusion — can flip one valid marker into another valid marker. The detector then reports a confident, wrong ID.

This is the core robustness limitation of CircularTargets, and it is why which markers you choose matters.

Measuring the family floor

You can reproduce the floor from the printed family alone:

  1. Decode each target in the Print Markers PDF to its N-bit integer (or read the published code tables — see Binary code tables below).
  2. For every pair, compute d(a, b) as the minimum Hamming distance over the N cyclic rotations of one code.
  3. The global minimum is the family floor; the per-pair values identify the closest (most confusable) pairs.

The pairwise distance is cheap to vectorize: for a 0/1 bit matrix B with per-row popcounts w, hamming(a, R*b) = w_a + w_b - 2*(B . (R*B)^T) for each rotation R, taking the elementwise minimum over the N rotations.

The masquerade (occlusion) effect

A subtler failure mode is masquerade: a partially occluded marker decoding as a different valid marker.

A CircularTarget is read by fitting the central dot, then sampling the surrounding ring segment by segment. When a marker sits in cluttered surroundings — a dark object overlapping one side, a bright highlight, debris — a contiguous arc of the ring can be forced to a single color (all black or all white) while the central dot stays visible and the target is still detected. If that occluded pattern happens to match another valid code (under the same rotation-invariant identification), the marker is silently read as a legitimate, different marker. When both markers belong to your project, the result is a confident wrong correspondence that is very hard to diagnose.

A pair {i, j} is forbidden when occluding some arc of i (any start, any end, either color) makes it identify as j, or vice versa. Because the CircularTarget floor is only 2, forbidden pairs are common:

Family Markers Forbidden pairs Share of all pairs
12-bit 161 1,652 12.8%
14-bit 561 9,414 6.0%
16-bit 2,001 49,523 2.5%
20-bit 26,013 1,174,354 0.35%

Smaller families are denser in code space, so relatively more occlusions land on another member. An unconstrained "most separated" set almost always still contains forbidden pairs — the separation objective and the occlusion objective are different — so masquerade avoidance has to be requested explicitly.

Detecting masquerade pairs

Identification reuses the decoder's own rotation-invariance, so no large rotation tables are needed:

  1. Reduce every marker's code to its canonical form — the minimum value over its N cyclic rotations — and build a lookup from canonical form to marker ID.
  2. For each color (black, white) and each contiguous arc (start, end) with wraparound, occlude every marker at once (code | arc_mask for white, code & ~arc_mask for black), canonicalize the result, and look it up.
  3. Wherever an occluded marker's canonical form matches a different marker, record the unordered forbidden pair.

Choosing a well-separated, occlusion-robust subset

Most projects do not need the whole library — they need k markers. "The first k" (targets 1...k) or a random subset is a poor choice: those subsets sit at or near the family floor, so they include the most confusable pairs. A deliberately chosen subset does markedly better at no practical cost.

The selection objective

The governing quantity is the worst (smallest) pairwise distance in the chosen set — the closest pair is what drives a mis-read. So the "best" k-subset is the one that maximizes the minimum pairwise distance (a max-min, or maximum-minimum-distance, objective). When the family is saturated and many subsets tie on the worst pair, a secondary objective — maximize the sum of pairwise distances — breaks the tie by spreading the picks out, minimizing how many pairs sit at the floor.

The algorithm, in brief

  • Greedy farthest-point (a 2-approximation). Seed with the two markers farthest apart, then repeatedly add the marker whose minimum distance to the already-chosen set is largest. This is the classic farthest-point heuristic (Gonzalez, 1985), within a factor of two of the optimal worst-pair distance; restarting from every seed and keeping the best result tightens it further.
  • Masquerade-free as a constraint. Treat the forbidden (masquerade) pairs as edges of a graph and require the chosen set to be an independent set in it (no forbidden pair). In the greedy, drop a candidate's forbidden neighbors as the set grows. This is nearly free: it holds the same worst-pair score and costs only a few hundredths of a bit of mean separation.
  • Exact optimum (small / medium families). The exact max-min for k markers equals the largest threshold t for which the graph of pairs at distance >= t contains a clique of size k. Solving maximum-clique on these threshold graphs (climbing up from the greedy lower bound, with a (k-1)-core prune) gives a proven optimum for the 12-, 14-, and 16-bit families. The 20-bit family is too large to solve exactly, so its figures are greedy lower bounds.
  • Largest masquerade-free pool. The largest subset with no forbidden pair is a maximum independent set in the forbidden graph, found exactly by integer linear programming (maximize the count subject to x_i + x_j <= 1 for every forbidden pair).

Which family for a given k?

More bits buy more separation. At k = 20, the best achievable worst-pair distance is:

Family Sector Library Best worst-pair (max-min) Normalized (/ N)
12-bit 30.0 deg 161 2 (saturated, exact) 0.17
14-bit 25.7 deg 561 4 (exact) 0.29
16-bit 22.5 deg 2,001 4 (exact) 0.25
20-bit 18.0 deg 26,013 >= 6 (greedy bound) 0.30

The 12-bit family is saturated at k = 20: even the optimal subset cannot exceed a worst pair of 2, so it cannot supply 20 mutually-robust markers. 14-bit lifts the worst pair to 4 and is the efficient default — near-best normalized separation with the largest sectors (so the smallest printable marker for a given pixels-per-sector budget). 16- and 20-bit give more absolute margin when your imaging can resolve the finer ring. Rule of thumb: pick the smallest N whose ring your imaging can resolve; among those, more bits means more separation. Detection and centroid precision depend on the marker's pixel size (the central-dot ellipse fit), not on bit count — more bits costs only ring-decoding resolution, not measurement accuracy.

Ready-to-use best 10-marker sets

The most-separated, masquerade-free 10-marker set per family. Worst-pair = the smallest pairwise rotation-invariant Hamming distance within the set (higher is more robust). Numbers are Metashape target numbers (the 1-based Print Markers label order):

Family Target numbers Worst-pair Mean
12-bit 13, 30, 49, 52, 81, 94, 123, 146, 153, 157 4 4.89
14-bit 31, 60, 63, 121, 206, 219, 378, 508, 549, 560 4 5.69
16-bit 81, 402, 430, 561, 853, 1177, 1691, 1929, 1996, 2000 4 7.02
20-bit 575, 2258, 9370, 13238, 20197, 21153, 24077, 25628, 25800, 26006 8* 8.84

* The 20-bit worst-pair is a greedy lower bound (the family is too large to verify exactly): treat "8" as ">= 6, probably 8" rather than a proven optimum. For comparison, a random 10-marker subset of the 12-bit family sits at worst-pair 2 (a confusable pair) versus 4 for the chosen set, and contains masquerade pairs that the chosen set does not.

Safer marker pools: the maximum masquerade-free sets

If you would rather pick your own markers than copy a fixed list, draw them from a maximum masquerade-free pool — the largest subset of a family that contains no forbidden (masquerade) pair at all. Any subset you then choose from such a pool is automatically free of occlusion masquerade between its members. You still want to spread your choices out for separation: these pools are saturated at worst-pair 2, so they remove masquerade, not the need to choose a well-separated subset. It is safer to pick your markers from these sets.

Each pool is the unique maximum independent set in its family's forbidden graph:

  • 12-bit: 79 of 161 markers are masquerade-free together.
  • 14-bit: 217 of 561 markers.
  • 16-bit: 809 of 2,001 markers.
12-bit — 79 masquerade-free target numbers 
7, 11, 13, 14, 15, 22, 26, 28, 29, 30, 34, 36, 37, 38, 39, 40, 41, 42, 43,
44, 49, 53, 54, 55, 56, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 70, 71,
72, 73, 74, 75, 76, 77, 78, 79, 81, 82, 83, 84, 85, 86, 88, 89, 95, 96,
97, 98, 99, 100, 101, 102, 103, 105, 106, 107, 108, 109, 110, 111, 112,
114, 115, 116, 117, 123, 124, 126, 127, 146
14-bit — 217 masquerade-free target numbers 
31, 62, 78, 85, 89, 91, 92, 93, 120, 134, 141, 145, 147, 148, 149, 161,
168, 172, 174, 175, 176, 183, 187, 189, 190, 191, 194, 196, 197, 198, 199,
200, 201, 202, 203, 204, 218, 230, 237, 241, 242, 243, 244, 254, 261, 265,
266, 267, 268, 274, 278, 279, 280, 281, 283, 284, 285, 286, 287, 288, 289,
290, 291, 292, 297, 304, 307, 308, 309, 310, 315, 318, 319, 320, 321, 323,
324, 325, 326, 327, 328, 329, 330, 331, 332, 334, 337, 338, 339, 340, 342,
343, 344, 345, 346, 347, 348, 349, 350, 351, 353, 354, 355, 356, 357, 358,
359, 360, 361, 362, 364, 365, 366, 367, 368, 370, 371, 381, 386, 388, 389,
390, 391, 394, 396, 397, 398, 399, 400, 401, 402, 403, 404, 405, 406, 407,
408, 412, 413, 414, 415, 416, 418, 419, 420, 421, 422, 423, 424, 425, 426,
427, 428, 429, 430, 431, 433, 434, 435, 436, 437, 438, 439, 440, 441, 442,
443, 444, 445, 447, 448, 449, 450, 451, 452, 453, 454, 456, 457, 458, 464,
465, 466, 467, 468, 469, 470, 471, 473, 474, 475, 476, 477, 478, 479, 480,
481, 482, 483, 485, 486, 487, 488, 489, 490, 492, 498, 500, 501, 502, 503,
505, 512, 530, 531, 532, 534
16-bit — 809 masquerade-free target numbers 
31, 47, 55, 59, 61, 62, 63, 94, 110, 118, 122, 124, 125, 126, 142, 150,
154, 156, 157, 158, 165, 169, 171, 172, 173, 177, 179, 180, 181, 183, 184,
185, 186, 187, 188, 217, 233, 240, 244, 246, 247, 248, 263, 270, 274, 276,
277, 278, 285, 289, 291, 292, 293, 297, 299, 300, 301, 303, 304, 305, 306,
307, 308, 322, 329, 333, 335, 336, 337, 344, 348, 350, 351, 352, 356, 358,
359, 360, 362, 363, 364, 365, 366, 367, 375, 379, 381, 382, 383, 387, 389,
390, 391, 393, 394, 395, 396, 397, 398, 402, 404, 405, 406, 408, 409, 410,
411, 412, 413, 415, 416, 417, 418, 419, 420, 422, 423, 424, 426, 444, 458,
465, 469, 471, 472, 473, 486, 493, 497, 499, 500, 501, 508, 512, 514, 515,
516, 520, 522, 523, 524, 525, 526, 527, 528, 529, 530, 541, 548, 552, 554,
555, 556, 563, 567, 569, 570, 571, 575, 577, 578, 579, 580, 581, 582, 583,
584, 585, 592, 596, 598, 599, 600, 604, 606, 607, 608, 609, 610, 611, 612,
613, 614, 617, 619, 620, 621, 622, 623, 624, 625, 626, 627, 629, 630, 631,
632, 633, 634, 636, 637, 638, 640, 647, 654, 658, 660, 661, 662, 669, 673,
675, 676, 677, 680, 682, 683, 684, 685, 686, 687, 688, 689, 690, 696, 700,
702, 703, 704, 707, 709, 710, 711, 712, 713, 714, 715, 716, 717, 720, 722,
723, 724, 725, 726, 727, 728, 729, 730, 732, 733, 734, 735, 736, 737, 739,
740, 741, 743, 747, 751, 753, 754, 755, 758, 760, 761, 762, 763, 764, 765,
766, 767, 768, 771, 773, 774, 775, 776, 777, 778, 779, 780, 781, 783, 784,
785, 786, 787, 788, 790, 791, 792, 794, 798, 800, 801, 802, 803, 804, 805,
806, 807, 808, 810, 811, 812, 813, 814, 815, 817, 818, 819, 821, 825, 826,
827, 828, 829, 831, 832, 833, 835, 839, 840, 842, 863, 870, 874, 875, 876,
877, 883, 887, 888, 889, 890, 893, 894, 895, 896, 897, 898, 899, 900, 901,
902, 910, 917, 921, 922, 923, 924, 930, 934, 935, 936, 937, 940, 941, 942,
943, 944, 945, 946, 947, 948, 949, 954, 958, 959, 960, 961, 964, 965, 966,
967, 968, 969, 970, 971, 972, 973, 975, 976, 977, 978, 979, 980, 981, 982,
983, 984, 986, 987, 988, 989, 990, 991, 993, 994, 995, 1003, 1007, 1008,
1009, 1010, 1016, 1020, 1021, 1022, 1023, 1025, 1026, 1027, 1028, 1029,
1030, 1031, 1032, 1033, 1034, 1038, 1042, 1043, 1044, 1045, 1047, 1048,
1049, 1050, 1051, 1052, 1053, 1054, 1055, 1056, 1058, 1059, 1060, 1061,
1062, 1063, 1064, 1065, 1066, 1067, 1069, 1070, 1071, 1072, 1073, 1074,
1076, 1077, 1078, 1083, 1084, 1085, 1086, 1088, 1089, 1090, 1091, 1092,
1093, 1094, 1095, 1096, 1097, 1099, 1100, 1101, 1102, 1103, 1104, 1105,
1106, 1107, 1108, 1110, 1111, 1112, 1113, 1114, 1115, 1117, 1118, 1119,
1124, 1125, 1126, 1127, 1128, 1129, 1130, 1131, 1132, 1134, 1135, 1136,
1137, 1138, 1139, 1141, 1142, 1143, 1148, 1149, 1150, 1151, 1153, 1154,
1155, 1160, 1173, 1174, 1175, 1176, 1177, 1178, 1179, 1180, 1181, 1182,
1185, 1188, 1189, 1190, 1191, 1193, 1194, 1195, 1196, 1197, 1198, 1199,
1200, 1201, 1202, 1204, 1205, 1206, 1207, 1208, 1209, 1210, 1211, 1212,
1213, 1215, 1216, 1217, 1218, 1219, 1220, 1222, 1223, 1228, 1229, 1230,
1232, 1233, 1234, 1235, 1236, 1237, 1238, 1239, 1240, 1241, 1243, 1244,
1245, 1246, 1247, 1248, 1249, 1250, 1251, 1252, 1254, 1255, 1256, 1257,
1258, 1259, 1261, 1262, 1267, 1268, 1269, 1270, 1271, 1272, 1273, 1274,
1276, 1277, 1278, 1279, 1280, 1281, 1283, 1284, 1289, 1290, 1291, 1293,
1294, 1299, 1311, 1312, 1313, 1315, 1316, 1317, 1318, 1319, 1320, 1321,
1322, 1323, 1324, 1326, 1327, 1328, 1329, 1330, 1332, 1333, 1338, 1339,
1340, 1341, 1342, 1343, 1344, 1346, 1347, 1348, 1349, 1350, 1352, 1353,
1358, 1359, 1360, 1362, 1363, 1368, 1381, 1382, 1383, 1384, 1385, 1387,
1388, 1393, 1394, 1395, 1397, 1398, 1403, 1416, 1417, 1422, 1458, 1460,
1461, 1462, 1463, 1464, 1465, 1466, 1467, 1468, 1469, 1470, 1471, 1472,
1475, 1476, 1477, 1478, 1479, 1480, 1481, 1482, 1483, 1484, 1485, 1486,
1487, 1489, 1490, 1491, 1492, 1493, 1494, 1495, 1496, 1498, 1499, 1500,
1501, 1503, 1508, 1509, 1510, 1511, 1512, 1513, 1514, 1515, 1516, 1518,
1519, 1520, 1521, 1522, 1523, 1524, 1525, 1527, 1528, 1529, 1530, 1532,
1537, 1538, 1539, 1540, 1541, 1543, 1544, 1545, 1546, 1548, 1553, 1555,
1570, 1571, 1572, 1573, 1574, 1575, 1576, 1577, 1578, 1580, 1581, 1582,
1583, 1584, 1585, 1586, 1587, 1589, 1590, 1591, 1592, 1598, 1599, 1600,
1601, 1602, 1603, 1604, 1605, 1607, 1608, 1609, 1610, 1611, 1612, 1613,
1614, 1616, 1617, 1618, 1619, 1625, 1626, 1627, 1628, 1630, 1631, 1632,
1633, 1654, 1655, 1657, 1658, 1659, 1665, 1666, 1667, 1669, 1670, 1671,
1742, 1743, 1745, 1746, 1766, 1772, 1774, 1929

Binary code tables (12-bit and 14-bit)

Each marker's identity is an N-bit integer. Bit s (value 2**s) corresponds to ring sector s, at angle s * 360/N degrees counter-clockwise from the 3-o'clock direction; a set bit is a black segment. By convention sector 0 is always black, so every code is odd. Because a printed target can be photographed at any orientation, the marker's identity is rotation-invariant: rotating the physical target cyclically shifts the bits, and the canonical identifier is the minimum value over the N cyclic rotations of the code.

Worked example (12-bit). Target #1 has code 71 = 0000 0100 0111 in 12 bits — sectors 0, 1, 2, and 6 are black. To recognize a rotated capture, compare canonical forms (the minimum over the 12 cyclic shifts) rather than the raw integer.

The integers below are listed in target-number order: the first value is target #1, the second is target #2, and so on.

12-bit — 161 codes (target order) 
71, 75, 77, 83, 85, 89, 95, 99, 101, 105, 111, 113, 119, 123, 125, 135,
139, 141, 147, 149, 153, 159, 163, 165, 169, 175, 177, 183, 187, 189, 195,
197, 201, 207, 209, 215, 219, 221, 231, 235, 237, 243, 245, 249, 255, 275,
277, 281, 287, 291, 293, 297, 303, 311, 315, 317, 325, 329, 335, 343, 347,
349, 359, 363, 365, 371, 373, 377, 383, 399, 407, 411, 413, 423, 427, 429,
435, 437, 441, 447, 455, 459, 461, 467, 469, 473, 479, 485, 489, 495, 503,
507, 509, 585, 591, 599, 603, 605, 615, 619, 621, 627, 629, 639, 663, 667,
669, 679, 683, 685, 691, 693, 703, 715, 717, 723, 725, 735, 751, 759, 763,
765, 819, 821, 831, 845, 853, 863, 879, 887, 891, 893, 927, 943, 951, 955,
957, 975, 983, 987, 989, 1003, 1005, 1013, 1023, 1365, 1375, 1391, 1399,
1403, 1455, 1463, 1467, 1495, 1499, 1535, 1755, 1791, 1919, 1983, 2015
14-bit — 561 codes (target order) 
135, 139, 141, 147, 149, 153, 159, 163, 165, 169, 175, 177, 183, 187, 189,
195, 197, 201, 207, 209, 215, 219, 221, 225, 231, 235, 237, 243, 245, 249,
255, 263, 267, 269, 275, 277, 281, 287, 291, 293, 297, 303, 305, 311, 315,
317, 323, 325, 329, 335, 337, 343, 347, 349, 353, 359, 363, 365, 371, 373,
377, 383, 387, 389, 393, 399, 401, 407, 411, 413, 417, 423, 427, 429, 435,
437, 441, 447, 455, 459, 461, 467, 469, 473, 479, 483, 485, 489, 495, 497,
503, 507, 509, 531, 533, 537, 543, 547, 549, 553, 559, 561, 567, 571, 573,
579, 581, 585, 591, 593, 599, 603, 605, 615, 619, 621, 627, 629, 633, 639,
645, 649, 655, 657, 663, 667, 669, 679, 683, 685, 691, 693, 697, 703, 711,
715, 717, 723, 725, 729, 735, 739, 741, 745, 751, 753, 759, 763, 765, 783,
785, 791, 795, 797, 807, 811, 813, 819, 821, 825, 831, 839, 843, 845, 851,
853, 857, 863, 867, 869, 873, 879, 881, 887, 891, 893, 903, 907, 909, 915,
917, 921, 927, 931, 933, 937, 943, 945, 951, 955, 957, 965, 969, 975, 977,
983, 987, 989, 999, 1003, 1005, 1011, 1013, 1017, 1023, 1093, 1097, 1103,
1111, 1115, 1117, 1127, 1131, 1133, 1139, 1141, 1145, 1151, 1161, 1167,
1175, 1179, 1181, 1191, 1195, 1197, 1203, 1205, 1209, 1215, 1223, 1227,
1229, 1235, 1237, 1241, 1247, 1251, 1253, 1257, 1263, 1271, 1275, 1277,
1303, 1307, 1309, 1319, 1323, 1325, 1331, 1333, 1337, 1343, 1351, 1355,
1357, 1363, 1365, 1369, 1375, 1379, 1381, 1385, 1391, 1399, 1403, 1405,
1419, 1421, 1427, 1429, 1433, 1439, 1443, 1445, 1449, 1455, 1463, 1467,
1469, 1481, 1487, 1495, 1499, 1501, 1511, 1515, 1517, 1523, 1525, 1529,
1535, 1587, 1589, 1593, 1599, 1607, 1611, 1613, 1619, 1621, 1625, 1631,
1637, 1641, 1647, 1655, 1659, 1661, 1677, 1683, 1685, 1689, 1695, 1701,
1705, 1711, 1719, 1723, 1725, 1737, 1743, 1751, 1755, 1757, 1767, 1771,
1773, 1779, 1781, 1785, 1791, 1823, 1829, 1833, 1839, 1847, 1851, 1853,
1865, 1871, 1879, 1883, 1885, 1895, 1899, 1901, 1907, 1909, 1913, 1919,
1935, 1943, 1947, 1949, 1959, 1963, 1965, 1971, 1973, 1977, 1983, 1995,
1997, 2003, 2005, 2009, 2015, 2021, 2025, 2031, 2039, 2043, 2045, 2343,
2347, 2349, 2355, 2357, 2367, 2379, 2381, 2387, 2389, 2399, 2405, 2415,
2423, 2427, 2429, 2451, 2453, 2463, 2469, 2479, 2487, 2491, 2493, 2511,
2519, 2523, 2525, 2535, 2539, 2541, 2547, 2549, 2559, 2643, 2645, 2655,
2671, 2679, 2683, 2685, 2709, 2719, 2735, 2743, 2747, 2749, 2767, 2775,
2779, 2781, 2791, 2795, 2797, 2803, 2805, 2815, 2863, 2871, 2875, 2877,
2895, 2903, 2907, 2909, 2919, 2923, 2925, 2931, 2933, 2943, 2967, 2971,
2973, 2983, 2987, 2989, 2995, 2997, 3007, 3021, 3027, 3029, 3039, 3055,
3063, 3067, 3069, 3279, 3287, 3291, 3293, 3303, 3307, 3309, 3317, 3327,
3383, 3387, 3389, 3407, 3415, 3419, 3421, 3431, 3435, 3437, 3445, 3455,
3483, 3485, 3495, 3499, 3501, 3509, 3519, 3541, 3551, 3567, 3575, 3579,
3581, 3701, 3711, 3741, 3755, 3757, 3765, 3775, 3797, 3807, 3823, 3831,
3835, 3837, 3903, 3925, 3935, 3951, 3959, 3963, 3965, 3999, 4015, 4023,
4027, 4029, 4055, 4059, 4061, 4075, 4077, 4085, 4095, 5463, 5467, 5483,
5503, 5547, 5567, 5599, 5615, 5623, 5627, 5823, 5855, 5871, 5879, 5883,
5983, 5999, 6007, 6011, 6063, 6071, 6075, 6107, 6143, 7023, 7031, 7095,
7167, 7679, 7935, 8063, 8127

Choosing between target families

Choice When Notes
12-bit circular Default. Single-chunk projects with a few dozen markers. Manual: "12 bit pattern is considered to be decoded more precisely."
14-bit circular Multi-chunk projects with up to a few hundred markers. Larger code library; precision slightly below 12-bit.
16-bit circular Industrial metrology with thousands of targets. Limited deployment in photogrammetry; mostly a metrology choice.
20-bit circular Massive code-library needs (huge surveyed sites). Rare in practice.
AprilTag (since 2.2) Robotics-derived workflows; needs robust detection in cluttered scenes. See AprilTag detection — choosing a variant.
Non-coded (white circle / 4-segment) Aerial / large-area projects where coded targets would be impractically large. Detected per-image automatically; identity assignment uses spatial position rather than encoded ID — match via ignore labels on import or rename detected markers post-detection.

For automation-friendly defaults, 12-bit circular targets + AprilTag (when 2.2+ is available) cover the vast majority of close-range and field-survey use cases.

Caveats

  • Library size limits. The 12-bit family has 161 unique codes; if your project needs more markers than that, you'll run out of unique IDs (causing identity collisions) and must switch to 14-bit (561 codes), 16-bit (2,001), or 20-bit (26,013). More important than raw count is separation: see "Choosing a well-separated, occlusion-robust subset" above — the 12-bit family cannot supply 20 mutually-robust markers even though it has 161 codes. Plan target counts and separation before printing.
  • Mounting flatness matters. A bent or curled target introduces ellipse-fitting bias in the central-circle detection. Mount targets on rigid panels (foamcore, aluminum, MDF) before deployment.
  • Lighting and contrast. Uniform lighting across the target's surface is critical. Targets in deep shadow have reduced contrast at the ring boundaries; the parity-check may then fail and the target is silently dropped from detection.
  • Surface reflectance. Glossy or reflective surfaces (laminated targets in direct sunlight) create specular highlights that confuse the detector. Matte vinyl or photo-quality matte paper outperforms glossy print.
  • Print accuracy. The central circle's diameter and the ring widths must be reproduced accurately by the printer. Inkjet printers can drift by a few percent; for metrology use, verify printed target dimensions with a caliper against the dimensions specified in the Print Markers dialog.
  • Aerial impracticality at high GSD. Per the manual, a 35-cm-diameter target for 1-cm-GSD aerial work scales to 175 cm at 5-cm GSD — impractical. Switch to non-coded targets (or pre-surveyed natural features) for high-GSD aerial work.
  • The 30-px upper bound is from a 2014-era tutorial. The detection algorithm has improved since; the upper bound is almost certainly looser today, but no updated guidance has been forum-published. Treat 30 px as a soft ceiling for conservative target sizing.
  • Patent status varies by jurisdiction. The German patent (Schneider 1991) has expired; equivalent patents in other jurisdictions may not have, depending on national filing dates. For commercial use, verify your local IP position before mass-producing 12-bit targets.

See also

References

  • Metashape Pro User Manual 2.3 (and Standard edition), p. 106–107, Working with coded and non-coded targets. The authoritative description of the four CircularTarget variants, the AprilTag family, the Print Markers workflow, and the precision / library-size tradeoff.
  • Metashape Python API Reference (2.3.1): Metashape.TargetType.CircularTarget12bit, CircularTarget14bit, CircularTarget16bit, CircularTarget20bit, plus the AprilTag variants; Chunk.detectMarkers(target_type=…).
  • Forum thread, The size of coded markers for UAV acquisition — the canonical sizing thread; a community-user reply (2017-04-03, PhotoScan 1.3) provides the empirical 9-10 px minimum central circle, 3.5× ratio for global diameter, and angle-of-view-secondary-to-effective-GSD finding.
  • Schneider, C. T. (1991). "3-D Vermessung von Oberflächen und Bauteilen durch Photogrammetrie und Bildverarbeitung." Proc. IDENT/VISION 91, 14–17. The original coded-target design publication; corresponding German patent has expired.
  • Photogrammetry targets: an overview (Petroff, 2018) — independent analysis of coded-target families in Metashape and other photogrammetry tools.
  • 14-Bit-Circular-Coded-Target (natowi, GitHub) — Python script that generates Metashape-compatible 14-bit circular coded targets outside of the Print Markers dialog.
  • CCTDecode (poxiao2, GitHub) — open-source decoder for circular coded targets in the same family as Metashape's CircularTarget.

  • Photoscan coded-target turntable JSFiddle (community) — generator for round-table target plates; replace the default count = Math.trunc(circ / (codeSize * 3)) line with count = Math.trunc(circ / (codeSize + padding)) for even target spacing on each ring.

  • Garrido-Jurado, S., et al. (2014). "Automatic generation and detection of highly reliable fiducial markers under occlusion." Pattern Recognition 47(6), 2280-2292. Background on occlusion-robust fiducial design and marker confusability.

  • Garrido-Jurado, S., et al. (2016). "Generation of fiducial marker dictionaries using mixed integer linear programming." Pattern Recognition 51, 481-491. Optimizing a marker dictionary for inter-marker distance.
  • Gonzalez, T. F. (1985). "Clustering to minimize the maximum intercluster distance." Theoretical Computer Science 38, 293-306. The farthest-point heuristic underlying max-min subset selection.
  • Coded targets and Scale bars (Agisoft KB) — Agisoft's basic workflow for printing, placing and auto-detecting circular coded targets and building scale bars from the detected markers.