相机模型封装

该模块提供了 CameraModel 类，用于封装 camera_models C++ 库（通过 pybind11 暴露），简化相机模型的加载和使用。

主要功能:

  • 在初始化时检查 camera_models 库，如果不存在则尝试使用 gettool 获取。

  • 从 YAML 文件加载相机参数。

  • 提供访问相机基本属性（如模型类型、名称、图像尺寸）的接口。

  • 封装核心的相机几何操作，如从 2D 图像点到 3D 射线的反投影 (lift_projective) 和从 3D 空间点到 2D 图像点的投影 (space_to_plane)。

  • 提供将相机参数导出为 Python 字典的方法。

主要方法和属性:

  • __init__(self): 初始化 CameraModel，检查并导入所需的 camera_models 库。

  • load_from_yaml(self, yaml_path): 从指定的 YAML 文件加载相机模型。

  • lift_projective(self, p): 将 2D 图像点 (Tuple, List, 或 np.ndarray) 提升到 3D projective 射线 (返回 np.ndarray)。

  • space_to_plane(self, P): 将 3D 空间点 (Tuple, List, 或 np.ndarray) 投影到 2D 图像平面 (返回 np.ndarray)。

  • get_parameters_as_dict(self): 返回包含相机详细参数的字典。

  • model_type (property): 获取相机模型类型。

  • camera_name (property): 获取相机名称。

  • image_width (property): 获取图像宽度。

  • image_height (property): 获取图像高度。

示例:

>>> from pywayne.cv.camera_model import CameraModel
>>> from pywayne.cv.tools import write_cv_yaml
>>> import os
>>> import numpy as np # 确保导入 numpy
>>>
>>> # 准备一个示例 YAML 文件 (实际应用中应从标定结果获取)
>>> sample_data = {
...     "model_type": "PINHOLE",
...     "camera_name": "my_cam",
...     "image_width": 1280,
...     "image_height": 720,
...     "distortion_parameters": {"k1": 0.0, "k2": 0.0, "p1": 0.0, "p2": 0.0},
...     "projection_parameters": {"fx": 600.0, "fy": 600.0, "cx": 640.0, "cy": 360.0}
... }
>>> yaml_file = "temp_cam_config.yaml"
>>> _ = write_cv_yaml(yaml_file, sample_data) # write_cv_yaml 返回布尔值
>>>
>>> # 初始化并加载模型
>>> cam = CameraModel()
>>> cam.load_from_yaml(yaml_file)
Loaded camera 'my_cam' (Type: ModelType.PINHOLE) from temp_cam_config.yaml # 示例输出可能略有不同
>>> print(f"Loaded camera: {cam.camera_name}, Type: {cam.model_type}")
Loaded camera: my_cam, Type: ModelType.PINHOLE
>>>
>>> # 获取参数
>>> params = cam.get_parameters_as_dict()
>>> print(f"Focal Length: fx={params['fx']:.1f}, fy={params['fy']:.1f}")
Focal Length: fx=600.0, fy=600.0
>>>
>>> # 投影和反投影示例
>>> image_point_list = [700.0, 400.0] # 使用 list 作为输入
>>> ray_3d_numpy = cam.lift_projective(image_point_list) # 返回 np.ndarray
>>> print(f"Lifted ray for {image_point_list}: {ray_3d_numpy} (type: {type(ray_3d_numpy)})")
Lifted ray for [700.0, 400.0]: [0.09901951 0.06601301 0.99278962] (type: <class 'numpy.ndarray'>)
>>>
>>> world_point_numpy = np.array([0.1, -0.2, 1.5]) # 使用 np.ndarray 作为输入
>>> projected_point_numpy = cam.space_to_plane(world_point_numpy) # 返回 np.ndarray
>>> print(f"Projected point for {world_point_numpy}: {projected_point_numpy} (type: {type(projected_point_numpy)})")
Projected point for [ 0.1 -0.2  1.5]: [680. 280.] (type: <class 'numpy.ndarray'>)
>>>
>>> # 清理临时文件
>>> os.remove(yaml_file)
class pywayne.cv.camera_model.CameraModel

Bases: object

A wrapper class for the camera_models C++ library exposed via pybind11. Handles library checking, loading models from YAML, and provides access to common camera operations and parameters.

property camera_name: str

The name of the loaded camera.

get_parameters_as_dict() → Dict[str, Any]

Returns the loaded camera parameters as a dictionary. Includes base parameters and attempts to include model-specific parameters if available.

Returns:

A dictionary containing the camera parameters.

property image_height: int

The height of the camera image in pixels.

property image_width: int

The width of the camera image in pixels.

lift_projective(p: Tuple[float, float] | List[float] | ndarray) → ndarray

Lifts a 2D image point to a 3D projective ray (unit vector).

Parameters:

p – A tuple, list, or NumPy array (u, v) representing the image point coordinates.

Returns:

A NumPy array (x, y, z) representing the 3D direction vector.

load_from_yaml(yaml_path: str | Path)

Loads a camera model from the specified YAML configuration file.

Parameters:

yaml_path – Path to the camera configuration YAML file.

Raises:

  • FileNotFoundError – If the YAML file does not exist.

  • RuntimeError – If the camera factory is not available or loading fails.

property model_type: Any

The model type of the loaded camera (e.g., PINHOLE, MEI).

space_to_plane(P: Tuple[float, float, float] | List[float] | ndarray) → ndarray

Projects a 3D point in camera coordinates onto the 2D image plane.

Parameters:

P – A tuple, list, or NumPy array (x, y, z) representing the 3D point.

Returns:

A NumPy array (u, v) representing the projected image point coordinates.