Merge branch 'subpixelapi' into 'subpixelapi' (c4a43a3d) · Commits · aflab / astrogeology / Autocnet

autocnet/matcher/subpixel.py

+36 −30

Original line number	Diff line number	Diff line
		@@ -1701,23 +1701,6 @@ def subpixel_register_point_smart(pointid,
		cost = None
		destination_node = nodes[measure.imageid]

		try:
		affine = estimate_local_affine(source_node.geodata,
		destination_node.geodata,
		source.apriorisample,
		source.aprioriline,
		30,
		30)
		except Exception as e:
		print(e)
		m = {'id': measure.id,
		'sample':measure.apriorisample,
		'line':measure.aprioriline,
		'status':False,
		'choosername':chooser}
		updated_measures.append([None, None, m])
		continue

		# Compute the baseline metrics using the smallest window
		size_x = np.inf
		size_y = np.inf
		@@ -1741,12 +1724,21 @@ def subpixel_register_point_smart(pointid,
		size_y=size_y,
		buffer=10)

		_, baseline_corr, _ = subpixel_template(reference_roi,
		moving_roi,
		affine=affine)
		try:
		baseline_affine = estimate_local_affine(reference_roi,
		moving_roi)
		except Exception as e:
		print(e)
		m = {'id': measure.id,
		'sample':measure.apriorisample,
		'line':measure.aprioriline,
		'status':False,
		'choosername':chooser}
		updated_measures.append([None, None, m])
		continue

		reference_clip = reference_roi.clip()
		moving_clip = moving_roi.clip(affine)
		moving_clip = moving_roi.clip(baseline_affine)
		if np.isnan(reference_clip).any() or np.isnan(moving_clip).any():
		print('Unable to process due to NaN values in the input data.')
		m = {'id': measure.id,
		@@ -1763,6 +1755,10 @@ def subpixel_register_point_smart(pointid,
		updated_measures.append([None, None, m])
		continue

		_, baseline_corr, _ = subpixel_template(reference_roi,
		moving_roi,
		affine=baseline_affine)

		baseline_mi = 0 #mutual_information_match(reference_roi, moving_roi, affine=affine)

		print(f'Baseline MI: {baseline_mi} \| Baseline Corr: {baseline_corr}')
		@@ -1783,7 +1779,17 @@ def subpixel_register_point_smart(pointid,
		size_y=match_kwargs['template_size'][1],
		buffer=10)


		try:
		affine = estimate_local_affine(reference_roi, moving_roi)
		except Exception as e:
		print(e)
		m = {'id': measure.id,
		'sample':measure.apriorisample,
		'line':measure.aprioriline,
		'status':False,
		'choosername':chooser}
		updated_measures.append([None, None, m])
		continue

		updated_affine, maxcorr, temp_corrmap = subpixel_template(reference_roi,
		moving_roi,
		@@ -2008,12 +2014,6 @@ def validate_candidate_measure(measure_to_register,
		line = measure_to_register['line']

		print(f'Validating measure: {measure_to_register_id} on image: {source_image.name}')
		try:
		affine = estimate_local_affine(source_node.geodata, destination_node.geodata, sample, line, 30,30)
		except:
		print('Unable to transform image to reference space. Likely too close to the edge of the non-reference image. Setting ignore=True')
		return [np.inf] * len(parameters)


		dists = []
		for parameter in parameters:
		@@ -2032,6 +2032,12 @@ def validate_candidate_measure(measure_to_register,
		size_y=match_kwargs['template_size'][1],
		buffer=10)

		try:
		affine = estimate_local_affine(reference_roi, moving_roi)
		except:
		print('Unable to transform image to reference space. Likely too close to the edge of the non-reference image. Setting ignore=True')
		return [np.inf] * len(parameters)

		updated_affine, maxcorr, temp_corrmap = subpixel_template(reference_roi,
		moving_roi,
		affine=affine)

autocnet/transformation/affine.py

+6 −19

Original line number	Diff line number	Diff line
		@@ -91,11 +91,10 @@ def estimate_affine_from_sensors(reference_image,
		return affine


		def estimate_local_affine(reference_image, moving_image, center_x, center_y, size_x, size_y):
		def estimate_local_affine(reference_roi, moving_roi):
		"""
		estimate_local_affine
		Applies the affine transfromation calculated in estimate_affine_from_sensors to the moving region of interest (ROI).

		similar to estimate_affine_from_sensors, but for regions of interest (ROI).

		Parameters
		----------
		@@ -105,16 +104,6 @@ def estimate_local_affine(reference_image, moving_image, center_x, center_y, siz
		moving_image : plio.io.io_gdal.GeoDataset
		Image that is expected to move around during the matching process,
		points are projected onto this image to compute an affine
		center_sample : number
		center x (aka center sample) of the ROI in reference image pixel space
		center_line : number
		center y (aka center line) of the ROI in reference image pixel space
		size_x : number
		distance from the center to the end of the ROI window in the x (aka sample) direction.
		This is in reference image pixel space, resulting roi shape is (sizey2, sizex2)
		size_y : number
		distance from the center to the end of the ROI window in the y (aka line) direction,
		This is in reference image pixel space, resulting roi shape is (sizey2, sizex2)

		Returns
		-------
		@@ -122,15 +111,13 @@ def estimate_local_affine(reference_image, moving_image, center_x, center_y, siz
		Affine matrix to transform the moving image onto the center image
		"""
		# get initial affine
		affine_transform = estimate_affine_from_sensors(reference_image, moving_image, center_x, center_y)

		ref_center = (center_x, center_y)
		affine_transform = estimate_affine_from_sensors(reference_roi.data, moving_roi.data, reference_roi.x, reference_roi.y)
		ref_center = (reference_roi.x, reference_roi.y)

		# MOVING NO AFFINE; Get the full moving image area so that an applied affine transformation that
		# adds no data around 1+ edge does not fool the to be applied matcher.
		# The affine transformed center is a better match than the a priori sensor coords at this point.
		affine_center = affine_transform(ref_center)[0]
		moving_roi = Roi(moving_image, *affine_center, size_x=size_x, size_y=size_y)

		# The above coordinate transformation to get the center of the ROI handles translation.
		# So, we only need to rotate/shear/scale the ROI. Omitting scale, which should be 1 (?) results

autocnet/transformation/roi.py

+6 −6

Original line number	Diff line number	Diff line
		@@ -143,7 +143,7 @@ class Roi():
		@property
		def center(self):
		ie = self.image_extent
		return ((ie[1] - ie[0])-1 + self.buffer2)/2. + 0.5, ((ie[3]-ie[2])-1 + self.buffer2)/2. + 0.5
		return ((ie[1] - ie[0])-1)/2. + 0.5, ((ie[3]-ie[2])-1)/2. + 0.5

		@property
		def is_valid(self):
		@@ -219,6 +219,10 @@ class Roi():
		mode=mode,
		order=3)

		if self.buffer != 0:
		pixel_locked = pixel_locked[self.buffer:-self.buffer,
		self.buffer:-self.buffer]

		if affine:
		# The cval is being set to the mean of the array,
		pixel_locked = tf.warp(pixel_locked,
		@@ -226,8 +230,4 @@ class Roi():
		order=3,
		mode=mode)

		if self.buffer != 0:
		return img_as_float32(pixel_locked[self.buffer:-self.buffer,
		self.buffer:-self.buffer])
		else:
		return img_as_float32(pixel_locked)