Single clean commit for parallel ground (#549) (96bc7794) · Commits · aflab / astrogeology / Autocnet

autocnet/cg/cg.py

+3 −3

Original line number	Diff line number	Diff line
		@@ -471,7 +471,7 @@ def distribute_points_in_geom(geom, method="classic",

		Returns
		-------
		valid : list
		valid : np.ndarray
		of valid points in the form (x,y) or (lon,lat)

		"""
		@@ -516,7 +516,7 @@ def distribute_points_in_geom(geom, method="classic",
		# Decision Tree
		if ratio < 0.16 and geom.area < 0.01:
		# Class: Slivers - ignore.
		return []
		return np.array([])
		elif geom.area <= 0.004 and ratio >= 0.25:
		# Single point at the centroid
		valid = single_centroid(geom)
		@@ -538,7 +538,7 @@ def distribute_points_in_geom(geom, method="classic",
		valid = point_distribution_func(geom, nspts, ewpts, Session=Session, **kwargs)
		else:
		print('WTF Willy')
		return valid
		return np.array(valid)


		def alpha_shape(points, alpha):

autocnet/graph/network.py

+116 −32

Original line number	Diff line number	Diff line
		@@ -44,8 +44,8 @@ from autocnet.graph.node import Node, NetworkNode
		from autocnet.io import network as io_network
		from autocnet.io.db import controlnetwork as io_controlnetwork
		from autocnet.io.db.model import (Images, Keypoints, Matches, Cameras, Points,
		Base, Overlay, Edges, Costs, Measures, JsonEncoder,
		try_db_creation)
		Base, Overlay, Edges, Costs, Measures, CandidateGroundPoints,
		JsonEncoder, try_db_creation)
		from autocnet.io.db.connection import new_connection, Parent
		from autocnet.matcher import subpixel
		from autocnet.matcher import cross_instrument_matcher as cim
		@@ -1350,7 +1350,10 @@ class NetworkCandidateGraph(CandidateGraph):
		'image': Images,
		'images': Images,
		'i': Images,
		5: Images
		5: Images,
		'candidategroundpoints' : CandidateGroundPoints,
		'candidategroundpoint' : CandidateGroundPoints,
		6: CandidateGroundPoints
		}

		def config_from_file(self, filepath):
		@@ -1583,6 +1586,7 @@ class NetworkCandidateGraph(CandidateGraph):
		reapply=False,
		log_dir=None,
		queue=None,
		exclude=None,
		**kwargs):
		"""
		A mirror of the apply function from the standard CandidateGraph object. This implementation
		@@ -1648,6 +1652,11 @@ class NetworkCandidateGraph(CandidateGraph):
		The processing queue to use. If None (default), use the processing queue from
		the config file.

		Returns
		-------
		job_str : str
		The string job that is submitted to the job scheduler

		Examples
		--------
		Apply a function to the overlay table omitting those overlay rows that already have
		@@ -1679,7 +1688,7 @@ class NetworkCandidateGraph(CandidateGraph):
		# Determine which obj will be called
		if isinstance(on, str):
		onobj = self.apply_iterable_options[on]
		elif isinstance(on, list):
		elif isinstance(on, (list, np.ndarray)):
		onobj = on

		# This method support arbitrary functions. The name needs to be a string for the log name.
		@@ -1691,7 +1700,7 @@ class NetworkCandidateGraph(CandidateGraph):
		# Dispatch to either the database object message generator or the autocnet object message generator
		if isinstance(onobj, DeclarativeMeta):
		job_counter = self._push_row_messages(onobj, on, function, walltime, filters, query_string, args, kwargs)
		elif isinstance(onobj, list):
		elif isinstance(onobj, (list, np.ndarray)):
		job_counter = self._push_iterable_message(onobj, function, walltime, args, kwargs)
		elif isinstance(onobj, (nx.classes.reportviews.EdgeView, nx.classes.reportviews.NodeView)):
		job_counter = self._push_obj_messages(onobj, function, walltime, args, kwargs)
		@@ -1723,8 +1732,10 @@ class NetworkCandidateGraph(CandidateGraph):
		time=walltime,
		partition=queue,
		output=log_dir+f'/autocnet.{function}-%j')
		submitter.submit(array='1-{}%{}'.format(job_counter,arraychunk), chunksize=chunksize)
		return job_counter
		job_str = submitter.submit(array='1-{}%{}'.format(job_counter,arraychunk),
		chunksize=chunksize,
		exclude=exclude)
		return job_str

		def generic_callback(self, msg):
		"""
		@@ -1783,7 +1794,8 @@ class NetworkCandidateGraph(CandidateGraph):

		Returns
		-------
		None
		df : pd.DataFrame
		The pandas dataframe that is passed to plio to generate the control network.

		"""
		# Read the cnet from the db
		@@ -1813,6 +1825,10 @@ class NetworkCandidateGraph(CandidateGraph):
		cnet.to_isis(df, path, targetname=target)
		cnet.write_filelist(fpaths, path=flistpath)

		# Even though this method writes, having a non-None return
		# let's a user work with the data that is passed to plio
		return df

		def update_from_jigsaw(self, path, pointid_func=lambda x: int(x.split('_')[-1])):
		"""
		Updates the measures table in the database with data from
		@@ -1919,11 +1935,17 @@ class NetworkCandidateGraph(CandidateGraph):
		----------
		img_path : str
		absolute path to image

		Returns
		-------
		node.id : int
		The id of the newly added node.
		"""
		image_name = os.path.basename(img_path)
		node = NetworkNode(image_path=img_path, image_name=image_name)
		node.parent = self
		node.populate_db()
		return node.id

		def copy_images(self, newdir):
		"""
		@@ -2381,11 +2403,92 @@ class NetworkCandidateGraph(CandidateGraph):
		submitter.submit(array='1-{}'.format(job_counter))
		return job_counter

		def distribute_ground_uniform(self, distribute_points_kwargs={}):
		"""
		Distribute candidate ground points into the union of the image footprints. This
		function returns a list of 2d nd-arrays where the first element is the longitude
		and the second element is the latitude.

		Parameters
		----------
		distirbute_points_kwargs : dict
		Of arguments that are passed on the the
		distribute_points_in_geom argument in autocnet.cg.cg
		Returns
		-------
		valid : np.ndarray
		n, 2 array with each row in the form lon, lat

		Examples
		--------
		To use this method, one can first define the spacing of ground points in the north-
		south and east-west directions using the `distribute_points_kwargs` keyword argument:
		def ns(x):
		from math import ceil
		return ceil(round(x,1)*3)
		def ew(x):
		from math import ceil
		return ceil(round(x,1)*3)

		Next these arguments can be passed in in order to generate the grid of points:
		distribute_points_kwargs = {'nspts_func':ns, 'ewpts_func':ew, 'method':'classic'}
		valid = ncg.distribute_ground_uniform(distribute_points_kwargs=distribute_points_kwargs)

		At this point, it is possible to visualize the valid points inside of a Jupyter notebook. This
		is frequently convenient when combined with the `ncg.union` property that displays the unioned
		geometries in the NetworkCandidateGraph.
		Finally, the valid points can be propagated using apply. The code below will use the defined base
		to find the most interesting ground feature in the region of the valid point and write that point
		to the table defined by CandidateGroundPoints (autocnet.io.db.model):

		base = 'mc11_oxia_palus_dir_final.cub'
		ncg.apply('matcher.ground.find_most_interesting_ground', on=valid, args=(base,))
		"""
		geom = self.union
		valid = cg.distribute_points_in_geom(geom, **distribute_points_kwargs)
		return valid

		def distribute_ground_density(self, threshold=4, distribute_points_kwargs={}):
		"""
		Distribute candidate ground points into overlaps with a number of images greater than or equal
		to the threshold. This function returns a list of 2d nd-arrays where the first element is the
		longitude and the second element is the latitude.

		Parameters
		----------
		distirbute_points_kwargs : dict
		Of arguments that are passed on the the
		distribute_points_in_geom argument in autocnet.cg.cg
		threshold : int
		Overlaps intersecting threshold images or greater have points placed.
		Default 4.
		Returns
		-------
		valid : np.ndarray
		n, 2 array in the form lon, lat

		Examples
		--------
		Usage for `distribute_ground_density` is identical to usage for `distribute_ground_uniform`.
		See Also
		--------
		autocnet.graph.network.NetworkCandidateGraph.distribute_ground_uniform
		"""
		valid = []
		with self.session_scope() as session:
		res = session.query(Overlay).filter(func.array_length(Overlay.intersections, 1) >= threshold).all()
		for r in res:
		coords = cg.distribute_points_in_geom(r.geom, **distribute_points_kwargs)
		if len(coords) > 0:
		valid.append(coords)
		valid = np.vstack(valid)
		return valid

		def subpixel_register_points(self, **kwargs):
		subpixel.subpixel_register_points(self.Session, **kwargs)

		def subpixel_register_point(self, pointid, **kwargs):
		subpixel.subpixel_register_point(self.Session, pointid, **kwarg)
		subpixel.subpixel_register_point(self.Session, pointid, **kwargs)

		def subpixel_regiter_mearure(self, measureid, **kwargs):
		subpixel.subpixel_register_measure(self.Session, measureid, **kwargs)
		@@ -2405,23 +2508,4 @@ class NetworkCandidateGraph(CandidateGraph):
		self.dem,
		nodes,
		**kwargs)
		def distribute_ground(self, distribute_points_kwargs={}):
		"""
		Distribute candidate ground points into the union of the image footprints. This
		function returns a list of 2d nd-arrays where the first element is the longitude
		and the second element is the latitude.

		Parameters
		----------
		distirbute_points_kwargs : dict
		Of arguments that are passed on the the
		distribute_points_in_geom argument in autocnet.cg.cg

		Returns
		-------
		valid : list
		of nd-arrays in the form [array([lon, lat]), array([lon, lat])]
		"""
		geom = self.union
		valid = cg.distribute_points_in_geom(geom, **distribute_points_kwargs)
		return valid

autocnet/io/db/model.py

+39 −6

Original line number	Diff line number	Diff line
		@@ -246,6 +246,7 @@ class Overlay(BaseMixin, Base):
		points = relationship('Points',
		primaryjoin='func.ST_Contains(foreign(Overlay.geom), Points.geom).as_comparison(1,2)',
		backref=backref('overlay', uselist=False),
		sync_backref=False,
		viewonly=True,
		uselist=True)

		@@ -303,7 +304,7 @@ class Points(BaseMixin, Base):
		ignore = Column("pointIgnore", Boolean, default=False)
		_apriori = Column("apriori", Geometry('POINTZ', srid=rectangular_srid, dimension=3, spatial_index=False))
		_adjusted = Column("adjusted", Geometry('POINTZ', srid=rectangular_srid, dimension=3, spatial_index=False))
		measures = relationship('Measures', back_populates="point")
		measures = relationship('Measures', order_by="asc(Measures.id)", back_populates="point")
		reference_index = Column("referenceIndex", Integer, default=0)

		@hybrid_property
		@@ -365,6 +366,39 @@ class Points(BaseMixin, Base):
		#def subpixel_register(self, Session, pointid, **kwargs):
		# subpixel.subpixel_register_point(args=(Session, pointid), **kwargs)


		class CandidateGroundPoints(BaseMixin, Base):
		__tablename__ = 'candidategroundpoints'
		latitudinal_srid = -1

		id = Column(Integer,primary_key=True, autoincrement=True)
		path = Column(String, nullable=False)
		choosername = Column("ChooserName", String)
		apriorisample = Column(Float)
		aprioriline = Column(Float)
		sample = Column(Float, nullable=False)
		line = Column(Float, nullable=False)
		_geom = Column("geom", Geometry('POINT', srid=latitudinal_srid, dimension=2, spatial_index=True))
		ignore = Column(Boolean, default=False)

		@hybrid_property
		def geom(self):
		try:
		return to_shape(self._geom)
		except:
		return self._geom

		@geom.setter
		def geom(self, newgeom):
		if isinstance(newgeom, osgeo.ogr.Geometry):
		# If an OGR geom, convert to shapely
		newgeom = shapely.wkt.loads(newgeom.ExportToWkt())
		if newgeom is None:
		self._geom = None
		else:
		self._geom = from_shape(newgeom, srid=self.latitudinal_srid)


		class MeasureType(enum.IntEnum):
		"""
		Enum to enforce measure type for ISIS control networks
		@@ -377,8 +411,8 @@ class MeasureType(enum.IntEnum):
		class Measures(BaseMixin, Base):
		__tablename__ = 'measures'
		id = Column(Integer,primary_key=True, autoincrement=True)
		pointid = Column(Integer, ForeignKey('points.id'), nullable=False)
		imageid = Column(Integer, ForeignKey('images.id'))
		pointid = Column(Integer, ForeignKey('points.id'), nullable=False, index=True)
		imageid = Column(Integer, ForeignKey('images.id'), index=True)
		serial = Column("serialnumber", String, nullable=False)
		_measuretype = Column("measureType", IntEnum(MeasureType), nullable=False) # [0,3] # Enum as above
		ignore = Column("measureIgnore", Boolean, default=False)
		@@ -443,7 +477,7 @@ def try_db_creation(engine, config):
		Points.rectangular_srid = rectangular_srid
		Points.semimajor_rad = spatial['semimajor_rad']
		Points.semiminor_rad = spatial['semiminor_rad']
		for cls in [Points, Overlay, Images, Keypoints, Matches]:
		for cls in [Points, Overlay, Images, Keypoints, Matches, CandidateGroundPoints]:
		setattr(cls, 'latitudinal_srid', latitudinal_srid)

		# If the table does not exist, this will create it. This is used in case a
		@@ -452,5 +486,4 @@ def try_db_creation(engine, config):
		Edges.__table__, Costs.__table__, Matches.__table__,
		Cameras.__table__, Points.__table__,
		Measures.__table__, Images.__table__,
		Keypoints.__table__])

		Keypoints.__table__, CandidateGroundPoints.__table__])
		No newline at end of file

autocnet/matcher/ground.py

0 → 100644

+353 −0

File added.

Preview size limit exceeded, changes collapsed.

autocnet/matcher/naive_template.py

+10 −27

Original line number	Diff line number	Diff line
		@@ -4,48 +4,40 @@ import numpy as np
		from scipy.ndimage.interpolation import zoom


		def pattern_match_autoreg(template, image, subpixel_size=3, max_scaler=0.2, func=cv2.TM_CCORR_NORMED):
		def pattern_match_autoreg(template, image, subpixel_size=3, max_scaler=0.2, metric=cv2.TM_CCORR_NORMED):
		"""
		Call an arbitrary pattern matcher using a subpixel approach where a center of gravity using
		the correlation coefficients are used for subpixel alignment.

		Parameters
		----------
		template : ndarray
		The input search template used to 'query' the destination
		image

		image : ndarray
		The image or sub-image to be searched

		subpixel_size : int
		An odd integer that defines the window size used to compute
		the moments

		max_scaler : float
		The percentage offset to apply to the delta between the maximum
		correlation and the maximum edge correlation.

		func : object
		metric : object
		The function to be used to perform the template based matching
		Options: {cv2.TM_CCORR_NORMED, cv2.TM_CCOEFF_NORMED, cv2.TM_SQDIFF_NORMED}
		In testing the first two options perform significantly better with Apollo data.

		Returns
		-------
		x : float
		The x offset

		y : float
		The y offset

		max_corr : float
		The strength of the correlation in the range [-1, 1].
		"""

		result = cv2.matchTemplate(image, template, method=func)
		result = cv2.matchTemplate(image, template, method=metric)

		if func == cv2.TM_SQDIFF or func == cv2.TM_SQDIFF_NORMED:
		if metric == cv2.TM_SQDIFF or metric == cv2.TM_SQDIFF_NORMED:
		y, x = np.unravel_index(np.argmin(result, axis=None), result.shape)
		else:
		y, x = np.unravel_index(np.argmax(result, axis=None), result.shape)
		@@ -59,7 +51,7 @@ def pattern_match_autoreg(template, image, subpixel_size=3, max_scaler=0.2, func

		if area.shape != (subpixel_size+2, subpixel_size+2):
		print("Max correlation is too close to the boundary.")
		return None, None, 0
		return None, None, 0, None

		# Find the max on the edges, scale just like autoreg (but why?)
		edge_max = np.max(np.vstack([area[0], area[-1], area[:,0], area[:,-1]]))
		@@ -83,43 +75,34 @@ def pattern_match_autoreg(template, image, subpixel_size=3, max_scaler=0.2, func
		y -= (image.shape[0] / 2) - (template.shape[0] / 2)
		x -= (image.shape[1] / 2) - (template.shape[1] / 2)

		return x, y, max_corr
		return float(x), float(y), float(max_corr), result

		def pattern_match(template, image, upsampling=16, metric=cv2.TM_CCOEFF_NORMED, error_check=False):
		"""
		Call an arbitrary pattern matcher using a subpixel approach where the template and image
		are upsampled using a third order polynomial.

		Parameters
		----------
		template : ndarray
		The input search template used to 'query' the destination
		image

		image : ndarray
		The image or sub-image to be searched

		upsampling : int
		The multiplier to upsample the template and image.

		func : object
		The function to be used to perform the template based matching
		Options: {cv2.TM_CCORR_NORMED, cv2.TM_CCOEFF_NORMED, cv2.TM_SQDIFF_NORMED}
		In testing the first two options perform significantly better with Apollo data.

		error_check : bool
		If True, also apply a different matcher and test that the values
		are not too divergent. Default, False.

		Returns
		-------

		x : float
		The x offset

		y : float
		The y offset

		strength : float
		The strength of the correlation in the range [-1, 1].
		"""