#include <camera_models.h>
Public Member Functions | |
| KannalaBrandtModel & | operator= (const KannalaBrandtModel &)=default |
| CameraIntrinsicsModel | GetType () const final |
| Getter for camera model type. | |
| int | NumberOfParameters () const final |
| Getter for the number of parameters for this camera model. | |
 Public Member Functions inherited from calico::sensors::CameraModel | |
| template<typename T > | |
| absl::StatusOr< Eigen::Vector2< T > > | ProjectPoint (const Eigen::VectorX< T > &intrinsics, const Eigen::Vector3< T > &point) const |
| template<typename T > | |
| absl::StatusOr< Eigen::Vector3< T > > | UnprojectPixel (const Eigen::VectorX< T > &intrinsics, const Eigen::Vector2< T > &pixel) const |
Static Public Member Functions | |
| template<typename T > | |
| static absl::StatusOr< Eigen::Vector2< T > > | ProjectPoint (const Eigen::VectorX< T > &intrinsics, const Eigen::Vector3< T > &point) |
| template<typename T > | |
| static absl::StatusOr< Eigen::Vector3< T > > | UnprojectPixel (const Eigen::VectorX< T > &intrinsics, const Eigen::Vector2< T > &pixel, int max_iterations=100) |
| Static Public Member Functions inherited from calico::sensors::CameraModel | |
| static std::unique_ptr< CameraModel > | Create (CameraIntrinsicsModel camera_model) |
Static Public Attributes | |
| static constexpr int | kNumberOfParameters = 7 |
| static constexpr CameraIntrinsicsModel | kModelType = CameraIntrinsicsModel::kKannalaBrandt |
Detailed Description
4-parameter Kannala-Brandt projection model as presented in OpenCV, also known as the "fisheye" model. This model assumes an isotropic pinhole model, i.e. \(f_x == f_y = f\).
Parameters are in the following order: \([f, c_x, c_y, k_1, k_2, k_3, k_4]\)
See https://docs.opencv.org/3.4/db/d58/group__calib3d__fisheye.html for more details.
Member Function Documentation
◆ ProjectPoint()
|
inlinestatic |
Returns projection \(\mathbf{p}\), a 2-D pixel coordinate such that
\[ \mathbf{p} = \left[\begin{matrix}f&0\\0&f\end{matrix}\right]\mathbf{p}_d + \left[\begin{matrix}c_x\\c_y\end{matrix}\right]\\ \mathbf{p}_d = \frac{\theta_d}{r}\mathbf{p}_m\\ \theta_d = \theta + k_1\theta^3 + k_2\theta^5 + k_3\theta^7 + k_4\theta^9\\ \theta = \arctan\left(r\right)\\ r^2 = {\mathbf{p}_m}^T\mathbf{p}_m\\ \mathbf{p}_m = \left[\begin{matrix}t_x / t_z\\t_y/t_z\end{matrix}\right]\\ \]
intrinsics is a vector of intrinsics parameters the following order: \([f, c_x, c_y, k_1, k_2, k_3, k_4]\)
point is the location of the feature resolved in the camera frame \(\mathbf{t}^s_{sx} = \left[\begin{matrix}t_x&t_y&t_z\end{matrix}\right]^T\).
◆ UnprojectPixel()
|
inlinestatic |
Inverts the projection model \(\mathbf{P}\) to obtain the bearing vector \(\mathbf{p}_m\) of the pixel location \(\mathbf{p}\). No closed form solution is available, so we use Newton's method to invert the projection model.
intrinsics is a vector of intrinsics parameters the following order: \([f, c_x, c_y, k_1, k_2, k_3, k_4]\)
max_iterations specifies the maximum number of Newton steps to take. Optimization will stop automatically if the error is less than 1e-14.
Note: This implementation seems to require a significant number of Newton steps to properly converge. If you need faster code, it is recommended that you use OpenCV's implementation which is better conditioned.
The documentation for this class was generated from the following file:
- calico/sensors/camera_models.h
