Rectification
This module provides functions to project an image onto a
real-world coordinate system using ground control points. The
module is largely based on the OpenCV Camera Calibration and 3D
Reconstruction workflow and works nicely together with the argus2
toolbox for coastal image analysis.
A typical workflow consists of determining ground control points
by measuring the real-world coordinates of object visible in the
image and the image coordinates of these very same objects. Also
the camera matrix and lens distortion parameters should be
determined.
Subsequently, a homography can be determined using the
flamingo.rectification.rectification.find_homography() function and a projection of the image can be
plotted using the accompanying flamingo.rectification.plot module.
Rectification
-
flamingo.rectification.rectification.find_homography(UV, XYZ, K, distortion=array([[ 0., 0., 0., 0.]]), z=0)[source]
Find homography based on ground control points
| Parameters: |
- UV (np.ndarray) – Nx2 array of image coordinates of gcp’s
- XYZ (np.ndarray) – Nx3 array of real-world coordinates of gcp’s
- K (np.ndarray) – 3x3 array containing camera matrix
- distortion (np.ndarray, optional) – 1xP array with distortion coefficients with P = 4, 5 or 8
- z (float, optional) – Real-world elevation on which the image should be projected
|
|---|
| Returns: | 3x3 homography matrix
|
|---|
| Return type: | np.ndarray
|
|---|
Notes
Function uses the OpenCV image rectification workflow as described in
http://docs.opencv.org/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html
starting with solvePnP.
Examples
>>> camera_id = 4
>>> r = argus2.rest.get_rectification_data('kijkduin')
>>> H = flamingo.rectification.find_homography(r[camera_id]['UV'],
r[camera_id]['XYZ'],
r[camera_id]['K'])
-
flamingo.rectification.rectification.get_pixel_coordinates(img)[source]
Get pixel coordinates given an image
| Parameters: | img (np.ndarray) – NxMx1 or NxMx3 image matrix |
|---|
| Returns: |
- np.ndarray –
NxM matrix containing u-coordinates
- np.ndarray –
NxM matrix containing v-coordinates
|
|---|
-
flamingo.rectification.rectification.rectify_coordinates(U, V, H)[source]
- Get projection of image pixels in real-world coordinates
- given image coordinate matrices and homography
| Parameters: |
- U (np.ndarray) – NxM matrix containing u-coordinates
- V (np.ndarray) – NxM matrix containing v-coordinates
- H (np.ndarray) – 3x3 homography matrix
|
|---|
| Returns: |
- np.ndarray –
NxM matrix containing real-world x-coordinates
- np.ndarray –
NxM matrix containing real-world y-coordinates
|
|---|
-
flamingo.rectification.rectification.rectify_image(img, H)[source]
- Get projection of image pixels in real-world coordinates
- given an image and homography
| Parameters: |
- img (np.ndarray) – NxMx1 or NxMx3 image matrix
- H (np.ndarray) – 3x3 homography matrix
|
|---|
| Returns: |
- np.ndarray –
NxM matrix containing real-world x-coordinates
- np.ndarray –
NxM matrix containing real-world y-coordinates
|
|---|
Visualization
-
flamingo.rectification.plot.find_horizon_offset(x, y, max_distance=10000.0)[source]
Find minimum number of pixels to crop to guarantee all pixels are within specified distance
| Parameters: |
- x (np.ndarray) – NxM matrix containing real-world x-coordinates
- y (np.ndarray) – NxM matrix containing real-world y-coordinates
- max_distance (float, optional) – Maximum distance from origin to be included in the plot.
Larger numbers are considered to be beyond the horizon.
|
|---|
| Returns: | Minimum crop distance in pixels (from the top of the image)
|
|---|
| Return type: | float
|
|---|
-
flamingo.rectification.plot.plot_coverage(X, Y, rotation=None, translation=None, max_distance=10000.0, ax=None, figsize=(30, 20), cmap=<matplotlib.colors.LinearSegmentedColormap object at 0x107b7de50>, alpha=0.4)[source]
Plot the coverage of the projection of multiple images in a single axis.
Plot the outline of lists of real-world x and y coordinate
matrices. The resulting composition can be rotated and
translated seperately.
Points projected at infinite distance can be ignored by
specifying a maximum distance.
| Parameters: |
- X (list of np.ndarrays) – List of NxM matrix containing real-world x-coordinates
- Y (list of np.ndarrays) – List of NxM matrix containing real-world y-coordinates
- rotation (float, optional) – Rotation angle in degrees
- translation (list or tuple, optional) – 2-tuple or list with x and y translation distances
- max_distance (float, optional) – Maximum distance from origin to be included in the plot.
Larger numbers are considered to be beyond the horizon.
- ax (matplotlib.axes.AxesSubplot, optional) – Axis object used for plotting
- figsize (tuple, optional) – 2-tuple or list containing figure dimensions
- cmap (matplotlib.colors.Colormap, optional) – Colormap to determine colors for individual patches
- alpha (float, optional) – Alpha value for patches
|
|---|
| Returns: |
- matplotlib.figure.Figure –
Figure object containing axis object
- matplotlib.axes.AxesSubplot –
Axis object containing plot
|
|---|
-
flamingo.rectification.plot.plot_rectified(X, Y, imgs, rotation=None, translation=None, max_distance=10000.0, ax=None, figsize=(30, 20), cmap='Greys', color=True, n_alpha=0)[source]
Plot the projection of multiple RGB images in a single axis.
Plot a list of images using corresponding lists of real-world
x and y coordinate matrices. The resulting composition can be
rotated and translated seperately.
Points projected at infinite distance can be ignored by
specifying a maximum distance.
| Parameters: |
- X (list of np.ndarrays) – List of NxM matrix containing real-world x-coordinates
- Y (list of np.ndarrays) – List of NxM matrix containing real-world y-coordinates
- imgs (list of np.ndarrays) – List of NxMx1 or NxMx3 image matrices
- rotation (float, optional) – Rotation angle in degrees
- translation (list or tuple, optional) – 2-tuple or list with x and y translation distances
- max_distance (float, optional) – Maximum distance from origin to be included in the plot.
Larger numbers are considered to be beyond the horizon.
- ax (matplotlib.axes.AxesSubplot, optional) – Axis object used for plotting
- figsize (tuple, optional) – 2-tuple or list containing figure dimensions
- color (bool, optional) – Whether color image should be plotted or grayscale
- n_alpha (int) – Number of border pixels to use to increase alpha
|
|---|
| Returns: |
- matplotlib.figure.Figure –
Figure object containing axis object
- matplotlib.axes.AxesSubplot –
Axis object containing plot
|
|---|
-
flamingo.rectification.plot.rotate_translate(x, y, rotation=None, translation=None)[source]
Rotate and/or translate coordinate system
| Parameters: |
- x (np.ndarray) – NxM matrix containing x-coordinates
- y (np.ndarray) – NxM matrix containing y-coordinates
- rotation (float, optional) – Rotation angle in degrees
- translation (list or tuple, optional) – 2-tuple or list with x and y translation distances
|
|---|
| Returns: |
- np.ndarrays –
NxM matrix containing rotated/translated x-coordinates
- np.ndarrays –
NxM matrix containing rotated/translated y-coordinates
|
|---|