Support SofastFixed point light #66
Conversation
braden6521
commented
Apr 2, 2024
braden6521
commented
- Updated the SofastFixed unit test data to use data captured with a point light near the center origin dot when using a printed, fixed pattern.
- Added more unit tests to the sofast/lib/image_processing library that contains the blob detection functions.
- Added an example that creates a measurement object from an image, finds the origin blob, and saves to HDF5 file.
braden6521
force-pushed
the
sofast_fixed_point_light
branch
from
13177e7 to
f4a7ed1
Compare
| @@ -422,8 +424,8 @@ def rectangle_loop_from_two_points(p1: Vxy, p2: Vxy, d_ax: float, d_perp: float) | |||
| # Calculate axial and perpendicular directions | |||
| v_axial = (p2 - p1).normalize() | |||
|
|
|||
| R = np.array([[0, 1], [-1, 0]]) | |||
| v_perp = v_axial.rotate(R) | |||
| rot_mat_90 = np.array([[0, 1], [-1, 0]]) | |||
| Parameters | ||
| ---------- | ||
| image : np.ndarray | ||
| Input image, uint8 |
There was a problem hiding this comment.
Presumably this should be a 2d image with a single color channel? AKA:
2D input image, single color channel, NxM or NxMx1, uint8
There was a problem hiding this comment.
Good idea! Changed!
| @@ -228,6 +235,50 @@ def test_calculate_active_pixel_pointing_vectors(self): | |||
| # Test | |||
| np.testing.assert_allclose(data['u_pixel_pointing_facet'], u_pixel_pointing_optic) | |||
|
|
|||
| def test_detect_blobs(self): | |||
| lt.logger(join(dir_save, 'log.txt'), lt.log.INFO) | ||
|
|
||
| # 1. Load image | ||
| # ============= |
There was a problem hiding this comment.
I hadn't considered also including this style of visual separator. Makes this really easy to read!
| # ============= | ||
| file_meas = join(opencsp_code_dir(), 'test/data/sofast_fixed/data_measurement/measurement_facet.h5') | ||
| measurement_old = MeasurementSofastFixed.load_from_hdf(file_meas) | ||
| image = measurement_old.image |
There was a problem hiding this comment.
I know it's silly, but it might make this more useful as an example if there were some commented out code for loading the image using opencv or pillow, as if you have an image jpg file instead of a measurement h5 file.
# image = cv2.imread(file_jpg, cv2.IMREAD_GRAYSCALE)There was a problem hiding this comment.
I like it! Added!
| # ================================ | ||
|
|
||
| # Find point light | ||
| params = cv.SimpleBlobDetector_Params() |
There was a problem hiding this comment.
Idea time! It would be nice if you could retrieve these parameters from image_processing with some values such as filterByCircularity set for you:
params = ip.dot_target_blob_params(minDistBetweenBlobs=2, minArea=3, maxArea=30)
def dot_target_blob_params(
minDistBetweenBlobs=2,
filterByArea=True,
minArea=3,
maxArea=30,
filterByCircularity=True,
minCircularity=0.8,
filterByConvexity=False,
filterByInertia=False,
):
params = cv.SimpleBlobDetector_Params()
params.minDistBetweenBlobs = minDistBetweenBlobs
params.filterByArea = filterByArea
params.minArea = minArea
params.maxArea = maxArea
params.filterByCircularity = filterByCircularity
params.minCircularity = minCircularity
params.filterByConvexity = filterByConvexity
params.filterByInertia = filterByInertia
return paramsTo be clear, not necessary for this PR, and not necessary generally. Just an idea.
There was a problem hiding this comment.
This is an interesting idea! There are other inputs that SimpleBlobDetector_Params takes that I haven't typically used. Perhaps have a default_blob_detector_params() function or something? I'll have to think about how best to implement it, but I do really like it. Will probably wait until a future PR.
| # ======================================= | ||
| image_annotated = ip.detect_blobs_inverse_annotate(image, params) | ||
|
|
||
| figure_control = rcfg.RenderControlFigure(tile_array=(1, 1), tile_square=True) |
There was a problem hiding this comment.
Unified render control style! 👍
|
@bbean23, thanks for the review! I implemented some of your suggestions, ready for re-review! |
