mirror of
https://github.com/opencv/opencv.git
synced 2026-07-25 21:33:04 +04:00
Siddharth Panditrao
0c613de006
Fix ~0.5px systematic offset in CharucoDetector subpixel refinement #28380 Resolves #25539 ### Problem `CharucoDetector::detectBoard` produces Charuco corner coordinates that are consistently offset by approximately **+0.5 pixels** compared to `findChessboardCorners + cornerSubPix`. This offset is systematic (mean ≈ −0.5 px when comparing legacy − Charuco) and reproducible across images. Visual inspection also shows that the legacy chessboard detector aligns better with the actual corner locations. ### Root cause In `charuco_detector.cpp`, the points passed to `cornerSubPix` are manually shifted by `-Point2f(0.5f, 0.5f)` before refinement and then shifted back by `+Point2f(0.5f, 0.5f)` after refinement. However, `cornerSubPix` refines corners in **absolute image coordinates** and converges to the true saddle point based on image gradients. Shifting the initial guess does not affect the converged result as long as the true corner lies within the refinement window. The additional `+0.5` shift applied after refinement therefore introduces a constant bias, resulting in: ``` P_out = P_true + 0.5 ``` ### Solution Remove the manual `±0.5` coordinate shifts around the `cornerSubPix` call and let the refined result be returned directly. ### Test updates Updated expected corner values in the following tests: - `testBoardSubpixelCoords` - `testSeveralBoardsWithCustomIds` The previous expected values (e.g. `200`, `250`, `300`) were only correct due to the +0.5px bias introduced by the bug. `generateImage` creates checkerboard squares of exactly **50 pixels**, which places true corner locations on **pixel boundaries** rather than pixel centers. As a result, the correct subpixel coordinates are: ``` 199.5, 249.5, 299.5 ``` instead of the previously expected integer values. After updating the expected values, all **30 Charuco-related tests pass** with the fix applied. ### Verification I verified the fix using a Python reproducer that compares `CharucoDetector` output against `findChessboardCorners + cornerSubPix`: - **Before fix:** mean error ≈ −0.501 px   - **After fix:** mean error ≈ −0.001 px   After the change, the Charuco detector output aligns with the legacy chessboard detector both numerically and visually. ### Notes This change only affects the post-refinement coordinate handling and does not alter detection logic, refinement parameters, or performance. See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [x] The feature is well documented and sample code can be built with the project CMake
249 lines
12 KiB
C++
249 lines
12 KiB
C++
// This file is part of OpenCV project.
|
|
// It is subject to the license terms in the LICENSE file found in the top-level directory
|
|
// of this distribution and at http://opencv.org/license.html.
|
|
|
|
#include "test_precomp.hpp"
|
|
#include "opencv2/objdetect/aruco_detector.hpp"
|
|
|
|
namespace opencv_test { namespace {
|
|
|
|
|
|
TEST(CV_ArucoTutorial, can_find_singlemarkersoriginal)
|
|
{
|
|
string img_path = cvtest::findDataFile("aruco/singlemarkersoriginal.jpg");
|
|
Mat image = imread(img_path);
|
|
aruco::ArucoDetector detector(aruco::getPredefinedDictionary(aruco::DICT_6X6_250));
|
|
|
|
vector<int> ids;
|
|
vector<vector<Point2f> > corners, rejected;
|
|
const size_t N = 6ull;
|
|
// corners of ArUco markers with indices goldCornersIds
|
|
const int goldCorners[N][8] = { {359,310, 404,310, 410,350, 362,350}, {427,255, 469,256, 477,289, 434,288},
|
|
{233,273, 190,273, 196,241, 237,241}, {298,185, 334,186, 335,212, 297,211},
|
|
{425,163, 430,186, 394,186, 390,162}, {195,155, 230,155, 227,178, 190,178} };
|
|
const int goldCornersIds[N] = { 40, 98, 62, 23, 124, 203};
|
|
map<int, const int*> mapGoldCorners;
|
|
for (size_t i = 0; i < N; i++)
|
|
mapGoldCorners[goldCornersIds[i]] = goldCorners[i];
|
|
|
|
detector.detectMarkers(image, corners, ids, rejected);
|
|
|
|
ASSERT_EQ(N, ids.size());
|
|
for (size_t i = 0; i < N; i++)
|
|
{
|
|
int arucoId = ids[i];
|
|
ASSERT_EQ(4ull, corners[i].size());
|
|
ASSERT_TRUE(mapGoldCorners.find(arucoId) != mapGoldCorners.end());
|
|
for (int j = 0; j < 4; j++)
|
|
{
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2]), corners[i][j].x, 1.f);
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2 + 1]), corners[i][j].y, 1.f);
|
|
}
|
|
}
|
|
}
|
|
|
|
TEST(CV_ArucoTutorial, can_find_gboriginal)
|
|
{
|
|
string imgPath = cvtest::findDataFile("aruco/gboriginal.jpg");
|
|
Mat image = imread(imgPath);
|
|
string dictPath = cvtest::findDataFile("aruco/tutorial_dict.yml");
|
|
aruco::Dictionary dictionary;
|
|
|
|
FileStorage fs(dictPath, FileStorage::READ);
|
|
dictionary.aruco::Dictionary::readDictionary(fs.root()); // set marker from tutorial_dict.yml
|
|
aruco::DetectorParameters detectorParams;
|
|
|
|
aruco::ArucoDetector detector(dictionary, detectorParams);
|
|
|
|
vector<int> ids;
|
|
vector<vector<Point2f> > corners, rejected;
|
|
const size_t N = 35ull;
|
|
// corners of ArUco markers with indices 0, 1, ..., 34
|
|
const int goldCorners[N][8] = { {252,74, 286,81, 274,102, 238,95}, {295,82, 330,89, 319,111, 282,104},
|
|
{338,91, 375,99, 365,121, 327,113}, {383,100, 421,107, 412,130, 374,123},
|
|
{429,109, 468,116, 461,139, 421,132}, {235,100, 270,108, 257,130, 220,122},
|
|
{279,109, 316,117, 304,140, 266,133}, {324,119, 362,126, 352,150, 313,143},
|
|
{371,128, 410,136, 400,161, 360,152}, {418,139, 459,145, 451,170, 410,163},
|
|
{216,128, 253,136, 239,161, 200,152}, {262,138, 300,146, 287,172, 248,164},
|
|
{309,148, 349,156, 337,183, 296,174}, {358,158, 398,167, 388,194, 346,185},
|
|
{407,169, 449,176, 440,205, 397,196}, {196,158, 235,168, 218,195, 179,185},
|
|
{243,170, 283,178, 269,206, 228,197}, {293,180, 334,190, 321,218, 279,209},
|
|
{343,192, 385,200, 374,230, 330,220}, {395,203, 438,211, 429,241, 384,233},
|
|
{174,192, 215,201, 197,231, 156,221}, {223,204, 265,213, 249,244, 207,234},
|
|
{275,215, 317,225, 303,257, 259,246}, {327,227, 371,238, 359,270, 313,259},
|
|
{381,240, 426,249, 416,282, 369,273}, {151,228, 193,238, 173,271, 130,260},
|
|
{202,241, 245,251, 228,285, 183,274}, {255,254, 300,264, 284,299, 238,288},
|
|
{310,267, 355,278, 342,314, 295,302}, {366,281, 413,290, 402,327, 353,317},
|
|
{125,267, 168,278, 147,314, 102,303}, {178,281, 223,293, 204,330, 157,317},
|
|
{233,296, 280,307, 263,346, 214,333}, {291,310, 338,322, 323,363, 274,349},
|
|
{349,325, 399,336, 386,378, 335,366} };
|
|
map<int, const int*> mapGoldCorners;
|
|
for (int i = 0; i < static_cast<int>(N); i++)
|
|
mapGoldCorners[i] = goldCorners[i];
|
|
|
|
detector.detectMarkers(image, corners, ids, rejected);
|
|
|
|
ASSERT_EQ(N, ids.size());
|
|
for (size_t i = 0; i < N; i++)
|
|
{
|
|
int arucoId = ids[i];
|
|
ASSERT_EQ(4ull, corners[i].size());
|
|
ASSERT_TRUE(mapGoldCorners.find(arucoId) != mapGoldCorners.end());
|
|
for (int j = 0; j < 4; j++)
|
|
{
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j*2]), corners[i][j].x, 1.f);
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j*2+1]), corners[i][j].y, 1.f);
|
|
}
|
|
}
|
|
}
|
|
|
|
TEST(CV_ArucoTutorial, can_find_choriginal)
|
|
{
|
|
string imgPath = cvtest::findDataFile("aruco/choriginal.jpg");
|
|
Mat image = imread(imgPath);
|
|
aruco::ArucoDetector detector(aruco::getPredefinedDictionary(aruco::DICT_6X6_250));
|
|
|
|
vector< int > ids;
|
|
vector< vector< Point2f > > corners, rejected;
|
|
const size_t N = 17ull;
|
|
// corners of aruco markers with indices goldCornersIds
|
|
const int goldCorners[N][8] = { {268,77, 290,80, 286,97, 263,94}, {360,90, 382,93, 379,111, 357,108},
|
|
{211,106, 233,109, 228,127, 205,123}, {306,120, 328,124, 325,142, 302,138},
|
|
{402,135, 425,139, 423,157, 400,154}, {247,152, 271,155, 267,174, 242,171},
|
|
{347,167, 371,171, 369,191, 344,187}, {185,185, 209,189, 203,210, 178,206},
|
|
{288,201, 313,206, 309,227, 284,223}, {393,218, 418,222, 416,245, 391,241},
|
|
{223,240, 250,244, 244,268, 217,263}, {333,258, 359,262, 356,286, 329,282},
|
|
{152,281, 179,285, 171,312, 143,307}, {267,300, 294,305, 289,331, 261,327},
|
|
{383,319, 410,324, 408,351, 380,347}, {194,347, 223,352, 216,382, 186,377},
|
|
{315,368, 345,373, 341,403, 310,398} };
|
|
map<int, const int*> mapGoldCorners;
|
|
for (int i = 0; i < static_cast<int>(N); i++)
|
|
mapGoldCorners[i] = goldCorners[i];
|
|
|
|
detector.detectMarkers(image, corners, ids, rejected);
|
|
|
|
ASSERT_EQ(N, ids.size());
|
|
for (size_t i = 0; i < N; i++)
|
|
{
|
|
int arucoId = ids[i];
|
|
ASSERT_EQ(4ull, corners[i].size());
|
|
ASSERT_TRUE(mapGoldCorners.find(arucoId) != mapGoldCorners.end());
|
|
for (int j = 0; j < 4; j++)
|
|
{
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2]), corners[i][j].x, 1.f);
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2 + 1]), corners[i][j].y, 1.f);
|
|
}
|
|
}
|
|
}
|
|
|
|
TEST(CV_ArucoTutorial, can_find_chocclusion)
|
|
{
|
|
string imgPath = cvtest::findDataFile("aruco/chocclusion_original.jpg");
|
|
Mat image = imread(imgPath);
|
|
aruco::ArucoDetector detector(aruco::getPredefinedDictionary(aruco::DICT_6X6_250));
|
|
|
|
vector< int > ids;
|
|
vector< vector< Point2f > > corners, rejected;
|
|
const size_t N = 13ull;
|
|
// corners of aruco markers with indices goldCornersIds
|
|
const int goldCorners[N][8] = { {301,57, 322,62, 317,79, 295,73}, {391,80, 413,85, 408,103, 386,97},
|
|
{242,79, 264,85, 256,102, 234,96}, {334,103, 357,109, 352,126, 329,121},
|
|
{428,129, 451,134, 448,152, 425,146}, {274,128, 296,134, 290,153, 266,147},
|
|
{371,154, 394,160, 390,180, 366,174}, {208,155, 232,161, 223,181, 199,175},
|
|
{309,182, 333,188, 327,209, 302,203}, {411,210, 436,216, 432,238, 407,231},
|
|
{241,212, 267,219, 258,242, 232,235}, {167,244, 194,252, 183,277, 156,269},
|
|
{202,314, 230,322, 220,349, 191,341} };
|
|
map<int, const int*> mapGoldCorners;
|
|
const int goldCornersIds[N] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15};
|
|
for (int i = 0; i < static_cast<int>(N); i++)
|
|
mapGoldCorners[goldCornersIds[i]] = goldCorners[i];
|
|
|
|
detector.detectMarkers(image, corners, ids, rejected);
|
|
|
|
ASSERT_EQ(N, ids.size());
|
|
for (size_t i = 0; i < N; i++)
|
|
{
|
|
int arucoId = ids[i];
|
|
ASSERT_EQ(4ull, corners[i].size());
|
|
ASSERT_TRUE(mapGoldCorners.find(arucoId) != mapGoldCorners.end());
|
|
for (int j = 0; j < 4; j++)
|
|
{
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2]), corners[i][j].x, 1.f);
|
|
EXPECT_NEAR(static_cast<float>(mapGoldCorners[arucoId][j * 2 + 1]), corners[i][j].y, 1.f);
|
|
}
|
|
}
|
|
}
|
|
|
|
TEST(CV_ArucoTutorial, can_find_diamondmarkers)
|
|
{
|
|
string imgPath = cvtest::findDataFile("aruco/diamondmarkers.jpg");
|
|
Mat image = imread(imgPath);
|
|
|
|
string dictPath = cvtest::findDataFile("aruco/tutorial_dict.yml");
|
|
aruco::Dictionary dictionary;
|
|
FileStorage fs(dictPath, FileStorage::READ);
|
|
dictionary.aruco::Dictionary::readDictionary(fs.root()); // set marker from tutorial_dict.yml
|
|
|
|
string detectorPath = cvtest::findDataFile("aruco/detector_params.yml");
|
|
fs = FileStorage(detectorPath, FileStorage::READ);
|
|
aruco::DetectorParameters detectorParams;
|
|
detectorParams.readDetectorParameters(fs.root());
|
|
detectorParams.cornerRefinementMethod = aruco::CORNER_REFINE_APRILTAG;
|
|
|
|
aruco::CharucoBoard charucoBoard(Size(3, 3), 0.4f, 0.25f, dictionary);
|
|
aruco::CharucoDetector detector(charucoBoard, aruco::CharucoParameters(), detectorParams);
|
|
|
|
vector<int> ids;
|
|
vector<vector<Point2f> > corners, diamondCorners;
|
|
vector<Vec4i> diamondIds;
|
|
const size_t N = 12ull;
|
|
// corner indices of ArUco markers
|
|
const int goldCornersIds[N] = { 4, 12, 11, 3, 12, 10, 12, 10, 10, 11, 2, 11 };
|
|
map<int, int> counterGoldCornersIds;
|
|
for (int i = 0; i < static_cast<int>(N); i++)
|
|
counterGoldCornersIds[goldCornersIds[i]]++;
|
|
|
|
const size_t diamondsN = 3;
|
|
// corners of diamonds with Vec4i indices
|
|
// Note: Values adjusted by -0.5px after fixing the systematic offset bug in charuco_detector.cpp
|
|
// The fix removes the incorrect +0.5 offset that was added after cornerSubPix
|
|
const float goldDiamondCorners[diamondsN][8] = {{195.1f,150.4f, 213.0f,200.7f, 135.9f,214.8f, 121.9f,163.0f},
|
|
{500.6f,170.8f, 501.4f,208.0f, 445.7f,199.3f, 447.3f,162.8f},
|
|
{342.9f,360.7f, 359.2f,328.2f, 400.3f,344.1f, 385.2f,377.9f}};
|
|
auto comp = [](const Vec4i& a, const Vec4i& b) {
|
|
for (int i = 0; i < 3; i++)
|
|
if (a[i] != b[i]) return a[i] < b[i];
|
|
return a[3] < b[3];
|
|
};
|
|
map<Vec4i, const float*, decltype(comp)> goldDiamonds(comp);
|
|
goldDiamonds[Vec4i(10, 4, 11, 12)] = goldDiamondCorners[0];
|
|
goldDiamonds[Vec4i(10, 3, 11, 12)] = goldDiamondCorners[1];
|
|
goldDiamonds[Vec4i(10, 2, 11, 12)] = goldDiamondCorners[2];
|
|
|
|
detector.detectDiamonds(image, diamondCorners, diamondIds, corners, ids);
|
|
map<int, int> counterRes;
|
|
|
|
ASSERT_EQ(N, ids.size());
|
|
for (size_t i = 0; i < N; i++)
|
|
{
|
|
int arucoId = ids[i];
|
|
counterRes[arucoId]++;
|
|
}
|
|
|
|
ASSERT_EQ(counterGoldCornersIds, counterRes); // check the number of ArUco markers
|
|
ASSERT_EQ(goldDiamonds.size(), diamondIds.size()); // check the number of diamonds
|
|
|
|
for (size_t i = 0; i < goldDiamonds.size(); i++)
|
|
{
|
|
Vec4i diamondId = diamondIds[i];
|
|
ASSERT_TRUE(goldDiamonds.find(diamondId) != goldDiamonds.end());
|
|
for (int j = 0; j < 4; j++)
|
|
{
|
|
EXPECT_NEAR(goldDiamonds[diamondId][j * 2], diamondCorners[i][j].x, 0.5f);
|
|
EXPECT_NEAR(goldDiamonds[diamondId][j * 2 + 1], diamondCorners[i][j].y, 0.5f);
|
|
}
|
|
}
|
|
}
|
|
|
|
}} // namespace
|