基于棋盘格的坐标转换

本教程展示了如何估计棋盘格的位姿并使用4x4齐次变换矩阵将点云转换到棋盘格坐标系。此示例还使用了转换矩阵来保存YAML文件。

本教程中使用的棋盘点云如下图所示。

我们可以在Zivid Studio中打开 原始点云 并进行检查。

备注

原始点云在 Sample Data（示例数据）中。

现在，我们可以将深度视图中的Z范围设置为540毫米至735毫米。这使我们可以看到相机和棋盘格坐标系之间存在一个角度。

首先，我们加载一个棋盘格的点云。

C++

const auto calibrationBoardFile = std::string(ZIVID_SAMPLE_DATA_DIR) + "/CalibrationBoardInCameraOrigin.zdf";
std::cout << "Reading ZDF frame from file: " << calibrationBoardFile << std::endl;
const auto frame = Zivid::Frame(calibrationBoardFile);
auto pointCloud = frame.pointCloud();

C＃

var calibrationBoardFile = Environment.GetFolderPath(Environment.SpecialFolder.CommonApplicationData)
               + "/Zivid/CalibrationBoardInCameraOrigin.zdf";
Console.WriteLine("Reading ZDF frame from file: " + calibrationBoardFile);
var frame = new Zivid.NET.Frame(calibrationBoardFile);
var pointCloud = frame.PointCloud;

Python

data_file = get_sample_data_path() / "CalibrationBoardInCameraOrigin.zdf"
print(f"Reading ZDF frame from file: {data_file}")
frame = zivid.Frame(data_file)
point_cloud = frame.point_cloud()

然后我们需要估计棋盘格的位姿。

C++

std::cout << "Detecting and estimating pose of the Zivid checkerboard in the camera frame" << std::endl;
const auto detectionResult = Zivid::Calibration::detectCalibrationBoard(frame);
const auto cameraToCheckerboardTransform = detectionResult.pose().toMatrix();

C＃

Console.WriteLine("Detecting and estimating pose of the Zivid checkerboard in the camera frame");
var detectionResult = Detector.DetectCalibrationBoard(frame);
var cameraToCheckerboardTransform = new Zivid.NET.Matrix4x4(detectionResult.Pose().ToMatrix());

Python

print("Detecting and estimating pose of the Zivid checkerboard in the camera frame")
detection_result = zivid.calibration.detect_calibration_board(frame)
camera_to_checkerboard_transform = detection_result.pose().to_matrix()

在转换点云之前，我们将变换矩阵求逆，以便得到相机在棋盘格坐标系中的位姿。

C++

std::cout << "Camera pose in checkerboard frame:" << std::endl;
const auto checkerboardToCameraTransform = cameraToCheckerboardTransform.inverse();

C＃

Console.WriteLine("Camera pose in checkerboard frame:");
var checkerboardToCameraTransform = cameraToCheckerboardTransform.Inverse();

Python

print("Camera pose in checkerboard frame:")
checkerboard_to_camera_transform = np.linalg.inv(camera_to_checkerboard_transform)

转换后，我们将位姿保存到YAML文件中。

C++

const auto transformFile = "CheckerboardToCameraTransform.yaml";
std::cout << "Saving a YAML file with Inverted checkerboard pose to file: " << transformFile << std::endl;
checkerboardToCameraTransform.save(transformFile);

C＃

var transformFile = "CheckerboardToCameraTransform.yaml";
Console.WriteLine("Saving a YAML file with Inverted checkerboard pose to file: " + transformFile);
checkerboardToCameraTransform.Save(transformFile);

Python

transform_file = Path("CheckerboardToCameraTransform.yaml")
print("Saving a YAML file with Inverted checkerboard pose to file: ")
assert_affine_matrix_and_save(checkerboard_to_camera_transform, transform_file)

这是YAML文件的内容：

__version__:
  serializer: 1
  data: 1
FloatMatrix:
  Data: [
    [0.9791644, 0.04366289, 0.1983198, 17.74656],
    [0.0502592, 0.8941201, -0.444998, 431.1943],
    [-0.1967516, 0.4456936, 0.8732962, -547.7883],
    [0, 0, 0, 1]]

之后，我们将点云转换到棋盘格坐标系。

C++

std::cout << "Transforming point cloud from camera frame to Checkerboard frame" << std::endl;
pointCloud.transform(checkerboardToCameraTransform);

C＃

Console.WriteLine("Transforming point cloud from camera frame to Checkerboard frame");
pointCloud.Transform(checkerboardToCameraTransform);

Python

print("Transforming point cloud from camera frame to Checkerboard frame")
point_cloud.transform(checkerboard_to_camera_transform)

提示

了解更多关于位置、方向和坐标变换。

在保存转换后的点云之前，我们可以将其转换为OpenCV 2D图像格式并绘制坐标系。

C++

std::cout << "Converting to OpenCV image format" << std::endl;
const auto bgraImage = pointCloudToColorBGRA(pointCloud);
std::cout << "Visualizing checkerboard with coordinate system" << std::endl;
drawCoordinateSystem(frame, cameraToCheckerboardTransform, bgraImage);
displayBGRA(bgraImage, "Checkerboard transformation frame");

Python

print("Converting to OpenCV image format")
bgra_image = point_cloud.copy_data("bgra")
print("Visualizing checkerboard with coordinate system")
_draw_coordinate_system(frame, camera_to_checkerboard_transform, bgra_image)
display_bgr(bgra_image, "Checkerboard transformation frame")

在这里我们可以看到将要显示的图像，我们可以观察到棋盘格的坐标系在哪里。

最后，我们将转换后的点云保存到磁盘。

C++

const auto checkerboardTransformedFile = "CalibrationBoardInCheckerboardOrigin.zdf";
std::cout << "Saving transformed point cloud to file: " << checkerboardTransformedFile << std::endl;
frame.save(checkerboardTransformedFile);

C＃

var checkerboardTransformedFile = "CalibrationBoardInCheckerboardOrigin.zdf";
Console.WriteLine("Saving transformed point cloud to file: " + checkerboardTransformedFile);
frame.Save(checkerboardTransformedFile);

Python

checkerboard_transformed_file = "CalibrationBoardInCheckerboardOrigin.zdf"
print(f"Saving transformed point cloud to file: {checkerboard_transformed_file}")
frame.save(checkerboard_transformed_file)

提示

了解更多关于位置、方向和坐标变换。

现在我们可以在Zivid Studio中打开 转换后的点云 并进行检查。

备注

在Zivid Studio中缩小图像来查找数据，因为视点原点不适用于转换后的点云。

我们现在可以在深度视图中手动将Z范围设置为-35mm至1mm，这样我们就可以过滤掉除了标定板和它旁边的物体之外的所有数据。这使我们可以看到我们在标定板上具有相同的Z值，并且从颜色渐变中我们可以检查该值是否为0。这意味着点云的原点位于棋盘格上。

要将点云转换至标定板坐标系，您可以运行我们的代码示例。

C++

示例： TransformPointCloudViaCheckerboard.cpp

./TransformPointCloudViaCheckerboard

C＃

示例： TransformPointCloudViaCheckerboard.cs

./TransformPointCloudViaCheckerboard

Python

示例： transform_point_cloud_via_checkerboard.py

python transform_point_cloud_via_checkerboard.py

小技巧

如果您希望在自己的设置中使用它，请修改代码示例：

用您的实际相机和设置替换 ZDF 文件。
将棋盘格标定板放置在场景中。
运行示例！