Source code for autocnet.matcher.cross_instrument_matcher

from plio.io.io_gdal import GeoDataset
import numpy as np
import os
import os.path

import pandas as pd

import geopandas as gpd

from geoalchemy2 import functions



from shapely import wkt
from shapely.geometry import Point

from autocnet.matcher.subpixel import check_match_func
from autocnet.io.db.model import Images, Points, Measures, JsonEncoder
from autocnet.cg.cg import distribute_points_in_geom, xy_in_polygon
from autocnet.spatial import isis
from autocnet.transformation.spatial import reproject, oc2og
from autocnet.matcher.cpu_extractor import extract_most_interesting
from autocnet.transformation import roi
from autocnet.matcher.subpixel import geom_match_simple
from autocnet.utils.utils import bytescale

import warnings

[docs]def generate_ground_points(Session, ground_mosaic, nspts_func=lambda x: int(round(x,1)*1), ewpts_func=lambda x: int(round(x,1)*4), size=(100,100)): """ Parameters ---------- ground_db_config : dict In the form: {'username':'somename', 'password':'somepassword', 'host':'somehost', 'pgbouncer_port':6543, 'name':'somename'} nspts_func : func describes distribution of points along the north-south edge of an overlap. ewpts_func : func describes distribution of points along the east-west edge of an overlap. size : tuple of int (size_x, size_y) maximum distances on either access point can move when attempting to find an interesting feature. """ if isinstance(ground_mosaic, str): ground_mosaic = GeoDataset(ground_mosaic) warnings.warn('This function is not well tested. No tests currently exist \ in the test suite for this version of the function.') session = Session() fp_poly = wkt.loads(session.query(functions.ST_AsText(functions.ST_Union(Images.geom))).one()[0]) session.close() coords = distribute_points_in_geom(fp_poly, nspts_func=nspts_func, ewpts_func=ewpts_func, method="new", Session=Session) coords = np.asarray(coords) old_coord_list = [] coord_list = [] lines = [] samples = [] newlines = [] newsamples = [] # throw out points not intersecting the ground reference images print('points to lay down: ', len(coords)) for i, coord in enumerate(coords): # res = ground_session.execute(formated_sql) p = Point(*coord) print(f'point {i}'), linessamples = isis.point_info(ground_mosaic.file_name, p.x, p.y, 'ground') if linessamples is None: print('unable to find point in ground image') continue line = linessamples.get('Line') sample = linessamples.get('Sample') oldpoint = isis.point_info(ground_mosaic.file_name, sample, line, 'image') op = Point(oldpoint.get('PositiveEast360Longitude'), oldpoint.get('PlanetocentricLatitude')) image = roi.Roi(ground_mosaic, sample, line, size_x=size[0], size_y=size[1]) image_roi = image.clip(dtype="uint64") interesting = extract_most_interesting(bytescale(image_roi), extractor_parameters={'nfeatures':30}) # kps are in the image space with upper left origin, so convert to # center origin and then convert back into full image space left_x, _, top_y, _ = image.image_extent newsample = left_x + interesting.x newline = top_y + interesting.y newpoint = isis.point_info(ground_mosaic.file_name, newsample, newline, 'image') p = Point(newpoint.get('PositiveEast360Longitude'), newpoint.get('PlanetocentricLatitude')) if not (xy_in_polygon(p.x, p.y, fp_poly)): print('Interesting point not in mosaic area, ignore') continue old_coord_list.append(op) lines.append(line) samples.append(sample) coord_list.append(p) newlines.append(newline) newsamples.append(newsample) # start building the cnet ground_cnet = pd.DataFrame() ground_cnet["path"] = [ground_mosaic.file_name]*len(coord_list) ground_cnet["pointid"] = list(range(len(coord_list))) ground_cnet["original point"] = old_coord_list ground_cnet["point"] = coord_list ground_cnet['original_line'] = lines ground_cnet['original_sample'] = samples ground_cnet['line'] = newlines ground_cnet['sample'] = newsamples ground_cnet = gpd.GeoDataFrame(ground_cnet, geometry='point') return ground_cnet, fp_poly, coord_list
[docs]def propagate_point(Session, config, dem, lon, lat, pointid, paths, lines, samples, size_x=40, size_y=40, match_func="classic", match_kwargs={'image_size': (39, 39), 'template_size': (21, 21)}, verbose=False, cost=lambda x, y: y == np.max(x)): """ Conditionally propagate a point into a stack of images. The point and all corresponding measures are matched against database network (to which you are propagating), best result(s) is(are) kept. Parameters ---------- Session : sqlalchemy.sessionmaker session maker associated with the database you want to propagate to config : dict configuation file associated with database you want to propagate to In the form: {'username':'somename', 'password':'somepassword', 'host':'somehost', 'pgbouncer_port':6543, 'name':'somename'} dem : surface surface model of target body lon : np.float longitude of point you want to project lat : np.float planetocentric latitude of point you want to project pointid : int clerical input used to trace point from generate_ground_points output paths : list of str absolute paths pointing to the image(s) from which you want to try porpagating the point lines : list of np.float apriori line(s) corresponding to point projected in 'paths' image(s) samples : list of np.float apriori sample(s) corresponding to point projected in 'paths' image(s) size_x : int half width of GeoDataset that is cut from full image and affinely transfromed in geom_match; must be larger than 1/2 template_kwargs 'image_size' size_y : int half height of GeoDataset that is cut from full image and affinely transfromed in geom_match; must be larger than 1/2 template_kwargs 'image_size' template_kwargs : dict kwargs passed through to control matcher.subpixel_template() verbose : boolean If True, this will print out the results of each propagation, including prints of the matcher areas and their correlation map. cost : anonymous function determines to which image(s) the point should be propagated. x corresponds to a list of all match correlation metrics, while y corresponds to each indiviudal element of the x array. Example: cost = lambda x,y: y == np.max(x) will get you one result corresponding to the image that has the maximum correlation with the source image cost = lambda x,y: y > 0.6 will propagate the point to all images whose correlation result is greater than 0.6 Returns ------- new_measures : pd.DataFrame Dataframe containing pointid, imageid, image serial number, line, sample, and ground location (both latlon and cartesian) of successfully propagated points """ match_func = check_match_func(match_func) session = Session() engine = session.get_bind() string = f"select * from images where ST_Intersects(geom, ST_SetSRID(ST_Point({lon}, {lat}), {config['spatial']['latitudinal_srid']}))" images = pd.read_sql(string, engine) session.close() image_measures = pd.DataFrame(zip(paths, lines, samples), columns=["path", "line", "sample"]) measure = image_measures.iloc[0] p = Point(lon, lat) new_measures = [] # list of matching results in the format: # [measure_index, x_offset, y_offset, offset_magnitude] match_results = [] # lazily iterate for now for k,m in image_measures.iterrows(): base_image = GeoDataset(m["path"]) sx, sy = m["sample"], m["line"] for i,image in images.iterrows(): # When grounding to THEMIS the df has a PATH to the QUAD dest_image = GeoDataset(image["path"]) if os.path.basename(m['path']) == os.path.basename(image['path']): continue try: print(f'prop point: base_image: {base_image}') print(f'prop point: dest_image: {dest_image}') print(f'prop point: (sx, sy): ({sx}, {sy})') x,y, dist, metrics, corrmap = geom_match_simple(base_image, dest_image, sx, sy, 16, 16, \ match_func = match_func, \ match_kwargs=match_kwargs, \ verbose=verbose) except Exception as e: raise Exception(e) match_results.append(e) continue match_results.append([k, x, y, metrics, dist, corrmap, m["path"], image["path"], image['id'], image['serial']]) # get best offsets match_results = np.asarray([res for res in match_results if isinstance(res, list) and all(r is not None for r in res)]) if match_results.shape[0] == 0: # no matches return new_measures # column index 3 is the metric returned by the geom matcher best_results = np.asarray([match for match in match_results if cost(match_results[:,3], match[3])]) if best_results.shape[0] == 0: # no matches satisfying cost return new_measures if verbose: print("match_results final length: ", len(match_results)) print("best_results length: ", len(best_results)) print("Full results: ", best_results) print("Winning CORRs: ", best_results[:,3], "Themis Pixel shifts: ", best_results[:,4]) print("Themis Images: ", best_results[:,6], "CTX images:", best_results[:,7]) print("Themis Sample: ", sx, "CTX Samples: ", best_results[:,1]) print("Themis Line: ", sy, "CTX Lines: ", best_results[:,2]) print('\n') # if the single best results metric (returned by geom_matcher) is None if len(best_results[:,3])==1 and best_results[:,3][0] is None: return new_measures height = dem.get_height(lat, lon) semi_major = config['spatial']['semimajor_rad'] semi_minor = config['spatial']['semiminor_rad'] # The CSM conversion makes the LLA/ECEF conversion explicit # reprojection takes ographic lat lon_og, lat_og = oc2og(lon, lat, semi_major, semi_minor) x, y, z = reproject([lon_og, lat_og, height], semi_major, semi_minor, 'latlon', 'geocent') for row in best_results: sample = row[1] line = row[2] new_measures.append({ 'pointid' : pointid, 'imageid' : row[8], 'serial' : row[9], 'path': row[7], 'line' : line, 'sample' : sample, 'template_metric' : row[3], 'template_shift' : row[4], 'point' : p, 'point_ecef' : Point(x, y, z) }) return new_measures
[docs]def propagate_control_network(Session, config, dem, base_cnet, size_x=40, size_y=40, match_func="classic", match_kwargs={'image_size': (39,39), 'template_size': (21,21)}, verbose=False, cost=lambda x,y: y == np.max(x)): """ Loops over a base control network's measure information (line, sample, image path) and uses image matching algorithms (autocnet.matcher.subpixel.geom_match) to find the corresponding line(s)/sample(s) in database images. Parameters ---------- Session : sqlalchemy.sessionmaker session maker associated with the database containing the images you want to propagate to config : dict configuation file associated with database containing the images you want to propagate to In the form: {'username':'somename', 'password':'somepassword', 'host':'somehost', 'pgbouncer_port':6543, 'name':'somename'} dem : surface surface model of target body base_cnet : pd.DataFrame Dataframe representing the points you want to propagate. Must contain 'line', 'sample' location of the measure and the 'path' to the corresponding image verbose : boolean Increase the level of print outs/plots recieved during propagation cost : anonymous function determines to which image(s) the point should be propagated. x corresponds to a list of all match correlation metrics, while y corresponds to each indiviudal element of the x array. Example: cost = lambda x,y: y == np.max(x) will get you one result corresponding to the image that has the maximum correlation with the source image cost = lambda x,y: y > 0.6 will propegate the point to all images whose correlation result is greater than 0.6 Returns ------- ground : pd.DataFrame Dataframe containing pointid, imageid, image serial number, line, sample, and ground location (both latlon and cartesian) of successfully propagated points """ warnings.warn('This function is not well tested. No tests currently exist \ in the test suite for this version of the function.') match_func = check_match_func(match_func) groups = base_cnet.groupby('pointid').groups # append CNET info into structured Python list constrained_net = [] # easily parallelized on the cpoint level, dummy serial for now for cpoint, indices in groups.items(): measures = base_cnet.loc[indices] measure = measures.iloc[0] p = measure.point # get image in the destination that overlap lon, lat = measures["point"].iloc[0].xy gp_measures = propagate_point(Session, config, dem, lon[0], lat[0], cpoint, measures["path"], measures["line"], measures["sample"], size_x, size_y, match_func, match_kwargs, verbose=verbose, cost=cost) # do not append if propagate_point is unable to find result if len(gp_measures) == 0: continue constrained_net.extend(gp_measures) ground = gpd.GeoDataFrame.from_dict(constrained_net).set_geometry('point') groundpoints = ground.groupby('pointid').groups points = [] # conditionally upload a new point to DB or updated existing point with new measures lat_srid = config['spatial']['latitudinal_srid'] session = Session() for gp_point, indices in groundpoints.items(): point = ground.loc[indices].iloc[0] # check DB for if point already exists there lon = point.point.x lat = point.point.y spatial_point = functions.ST_Point(lon, lat) spatial_setSRID = functions.ST_SetSRID(spatial_point, lat_srid) spatial_buffer = functions.ST_Buffer(spatial_setSRID, 10e-10) spatial_intersects = functions.ST_Intersects(Points.geom, spatial_buffer) res = session.query(Points).filter(spatial_intersects).all() if len(res) > 1: warnings.warn(f"There is more than one point at lon: {lon}, lat: {lat}") elif len(res) == 1: # update existing point with new measures for i in indices: row = ground.loc[i] pid = res[0].id meas = session.query(Measures.serial).filter(Measures.pointid == pid).all() serialnumbers = [m[0] for m in meas] if row['serial'] in serialnumbers: continue points.append(Measures(pointid = pid, line = float(row['line']), sample = float(row['sample']), aprioriline = float(row['line']), apriorisample = float(row['sample']), imageid = int(row['imageid']), template_metric = float(row['template_metric']), template_shift = float(row['template_shift']), serial = row['serial'], measuretype = 3)) else: # upload new point p = Points() p.pointtype = 3 p.apriori = point['point_ecef'] p.adjusted = point['point_ecef'] for i in indices: row = ground.loc[i] p.measures.append(Measures(line = float(row['line']), sample = float(row['sample']), aprioriline = float(row['line']), apriorisample = float(row['sample']), imageid = int(row['imageid']), serial = row['serial'], template_metric = float(row['template_metric']), template_shift = float(row['template_shift']), measuretype = 3)) points.append(p) session.add_all(points) session.commit() session.close() return ground