import cv2 import pyexiv2 import re from fractions import Fraction from opensfm.exif import sensor_string import log import io import system import context class ODM_Photo: """ ODMPhoto - a class for ODMPhotos """ def __init__(self, path_file, force_focal, force_ccd): # general purpose self.path_file = path_file self.filename = io.extract_file_from_path_file(path_file) # useful attibutes self.width = None self.height = None self.ccd_width = None self.focal_length = None self.focal_length_px = None # other attributes self.camera_make = '' self.camera_model = '' self.make_model = '' # parse values from metadata self.parse_pyexiv2_values(self.path_file, force_focal, force_ccd) # compute focal length into pixels self.update_focal() # print log message log.ODM_DEBUG('Loaded %s | camera: %s | dimensions: %s x %s | focal: %s | ccd: %s' % (self.filename, self.make_model, self.width, self.height, self.focal_length, self.ccd_width)) def update_focal(self): # compute focal length in pixels if self.focal_length and self.ccd_width: # take width or height as reference if self.width > self.height: # f(px) = w(px) * f(mm) / ccd(mm) self.focal_length_px = \ self.width * (self.focal_length / self.ccd_width) else: # f(px) = h(px) * f(mm) / ccd(mm) self.focal_length_px = \ self.height * (self.focal_length / self.ccd_width) def parse_pyexiv2_values(self, _path_file, _force_focal, _force_ccd): # read image metadata metadata = pyexiv2.ImageMetadata(_path_file) metadata.read() # loop over image tags for key in metadata: # try/catch tag value due to weird bug in pyexiv2 # ValueError: invalid literal for int() with base 10: '' try: # parse tag names if key == 'Exif.Image.Make': self.camera_make = metadata[key].value elif key == 'Exif.Image.Model': self.camera_model = metadata[key].value elif key == 'Exif.Photo.FocalLength': self.focal_length = float(metadata[key].value) except (pyexiv2.ExifValueError, ValueError) as e: pass except NotImplementedError as e: pass if self.camera_make and self.camera_model: self.make_model = sensor_string(self.camera_make, self.camera_model) # needed to do that since sometimes metadata contains wrong data img = cv2.imread(_path_file) self.width = img.shape[1] self.height = img.shape[0] # force focal and ccd_width with user parameter if _force_focal: self.focal_length = _force_focal if _force_ccd: self.ccd_width = _force_ccd # find ccd_width from file if needed if self.ccd_width is None and self.camera_model is not None: # load ccd_widths from file ccd_widths = system.get_ccd_widths() # search ccd by camera model key = [x for x in ccd_widths.keys() if self.make_model in x] # convert to float if found if key: self.ccd_width = float(ccd_widths[key[0]]) else: log.ODM_WARNING('Could not find ccd_width in file. Use --force-ccd or edit the sensor_data.json ' 'file to manually input ccd width') # TODO: finish this class class ODM_Reconstruction(object): """docstring for ODMReconstruction""" def __init__(self, arg): super(ODMReconstruction, self).__init__() self.arg = arg class ODM_GCPoint(object): """docstring for ODMPoint""" def __init__(self, x, y, z): self.x = x self.y = y self.z = z class ODM_GeoRef(object): """docstring for ODMUtmZone""" def __init__(self): self.datum = 'WGS84' self.epsg = None self.utm_zone = 0 self.utm_pole = 'N' self.utm_east_offset = 0 self.utm_north_offset = 0 self.gcps = [] def calculate_EPSG(self, _utm_zone, _pole): """Calculate and return the EPSG""" if _pole == 'S': return 32700 + _utm_zone elif _pole == 'N': return 32600 + _utm_zone else: log.ODM_ERROR('Unknown pole format %s' % _pole) return def coord_to_fractions(self, coord, refs): deg_dec = abs(float(coord)) deg = int(deg_dec) minute_dec = (deg_dec - deg) * 60 minute = int(minute_dec) sec_dec = (minute_dec - minute) * 60 sec_dec = round(sec_dec, 3) sec_denominator = 1000 sec_numerator = int(sec_dec * sec_denominator) if float(coord) >= 0: latRef = refs[0] else: latRef = refs[1] output = str(deg) + '/1 ' + str(minute) + '/1 ' + str(sec_numerator) + '/' + str(sec_denominator) return output, latRef def convert_to_las(self, _file, _file_out, json_file): if not self.epsg: log.ODM_ERROR('Empty EPSG: Could not convert to LAS') return kwargs = {'bin': context.pdal_path, 'f_in': _file, 'f_out': _file_out, 'east': self.utm_east_offset, 'north': self.utm_north_offset, 'epsg': self.epsg, 'json': json_file} # create pipeline file transform.xml to enable transformation pipeline = '{{' \ ' "pipeline":[' \ ' "untransformed.ply",' \ ' {{' \ ' "type":"filters.transformation",' \ ' "matrix":"1 0 0 {east} 0 1 0 {north} 0 0 1 0 0 0 0 1"' \ ' }},' \ ' {{' \ ' "a_srs":"EPSG:{epsg}",' \ ' "offset_x":"{east}",' \ ' "offset_y":"{north}",' \ ' "offset_z":"0",' \ ' "filename":"transformed.las"' \ ' }}' \ ' ]' \ '}}'.format(**kwargs) with open(json_file, 'w') as f: f.write(pipeline) # call pdal system.run('{bin}/pdal pipeline -i {json} --readers.ply.filename={f_in} ' '--writers.las.filename={f_out}'.format(**kwargs)) def convert_to_dem(self, _file, _file_out, sample_radius, gdal_res, gdal_radius): # Check if exists f_in if not io.file_exists(_file): log.ODM_ERROR('LAS file does not exist') return False kwargs = { 'bin': context.pdal_path, 'f_in': _file, 'sample_radius': sample_radius, 'gdal_res': gdal_res, 'gdal_radius': gdal_radius, 'f_out': _file_out } system.run('{bin}/pdal pipeline {json} --readers.las.filename={f_in} ' '--writers.gdal.filename={f_out}'.format(**kwargs)) if io.file_exists(kwargs['f_out']): return True else: return False def utm_to_latlon(self, _file, _photo, idx): gcp = self.gcps[idx] kwargs = {'epsg': self.epsg, 'file': _file, 'x': gcp.x + self.utm_east_offset, 'y': gcp.y + self.utm_north_offset, 'z': gcp.z} latlon = system.run_and_return('echo {x} {y} {z} '.format(**kwargs), 'gdaltransform -s_srs \"EPSG:{epsg}\" ' '-t_srs \"EPSG:4326\"'.format(**kwargs)).split() # Example: 83d18'16.285"W # Example: 41d2'11.789"N # Example: 0.998 if len(latlon) == 3: lon_str, lat_str, alt_str = latlon elif len(latlon) == 2: lon_str, lat_str = latlon alt_str = '' else: log.ODM_ERROR('Something went wrong %s' % latlon) lat_frac = self.coord_to_fractions(latlon[1], ['N', 'S']) lon_frac = self.coord_to_fractions(latlon[0], ['E', 'W']) # read image metadata metadata = pyexiv2.ImageMetadata(_photo.path_file) metadata.read() # set values # GPS latitude key = 'Exif.GPSInfo.GPSLatitude' value = lat_frac[0].split(' ') log.ODM_DEBUG('lat_frac: %s %s %s' % (value[0], value[1], value[2])) metadata[key] = pyexiv2.ExifTag(key, [Fraction(value[0]), Fraction(value[1]), Fraction(value[2])]) key = 'Exif.GPSInfo.GPSLatitudeRef' value = lat_frac[1] metadata[key] = pyexiv2.ExifTag(key, value) # GPS longitude key = 'Exif.GPSInfo.GPSLongitude' value = lon_frac[0].split(' ') metadata[key] = pyexiv2.ExifTag(key, [Fraction(value[0]), Fraction(value[1]), Fraction(value[2])]) key = 'Exif.GPSInfo.GPSLongitudeRef' value = lon_frac[1] metadata[key] = pyexiv2.ExifTag(key, value) # GPS altitude altitude = abs(int(float(latlon[2]) * 100)) key = 'Exif.GPSInfo.GPSAltitude' value = Fraction(altitude, 1) metadata[key] = pyexiv2.ExifTag(key, value) if latlon[2] >= 0: altref = '0' else: altref = '1' key = 'Exif.GPSInfo.GPSAltitudeRef' metadata[key] = pyexiv2.ExifTag(key, altref) # write values metadata.write() def parse_coordinate_system(self, _file): """Write attributes to jobOptions from coord file""" # check for coordinate file existence if not io.file_exists(_file): log.ODM_ERROR('Could not find file %s' % _file) return with open(_file) as f: # extract reference system and utm zone from first line. # We will assume the following format: # 'WGS84 UTM 17N' line = f.readline() log.ODM_DEBUG('Line: %s' % line) ref = line.split(' ') # match_wgs_utm = re.search('WGS84 UTM (\d{1,2})(N|S)', line, re.I) if ref[0] == 'WGS84' and ref[1] == 'UTM': # match_wgs_utm: self.datum = ref[0] self.utm_pole = ref[2][len(ref) - 1] self.utm_zone = int(ref[2][:len(ref) - 1]) # extract east and west offsets from second line. # We will assume the following format: # '440143 4588391' # update EPSG self.epsg = self.calculate_EPSG(self.utm_zone, self.utm_pole) # If the first line looks like "EPSG:n" or "epsg:n" elif ref[0].split(':')[0].lower() == 'epsg': self.epsg = line.split(':')[1] else: log.ODM_ERROR('Could not parse coordinates. Bad CRS supplied: %s' % line) return offsets = f.readline().split(' ') self.utm_east_offset = int(offsets[0]) self.utm_north_offset = int(offsets[1]) # parse coordinates lines = f.readlines() for l in lines: xyz = l.split(' ') if len(xyz) == 3: x, y, z = xyz[:3] elif len(xyz) == 2: x, y = xyz[:2] z = 0 self.gcps.append(ODM_GCPoint(float(x), float(y), float(z))) # Write to json file class ODM_Tree(object): def __init__(self, root_path, images_path): # root path to the project self.root_path = io.absolute_path_file(root_path) if not images_path: self.input_images = io.join_paths(self.root_path, 'images') else: self.input_images = io.absolute_path_file(images_path) # modules paths # here are defined where all modules should be located in # order to keep track all files al directories during the # whole reconstruction process. self.dataset_raw = io.join_paths(self.root_path, 'images') self.dataset_resize = io.join_paths(self.root_path, 'images_resize') self.opensfm = io.join_paths(self.root_path, 'opensfm') self.pmvs = io.join_paths(self.root_path, 'pmvs') self.odm_meshing = io.join_paths(self.root_path, 'odm_meshing') self.odm_texturing = io.join_paths(self.root_path, 'odm_texturing') self.odm_25dtexturing = io.join_paths(self.root_path, 'odm_texturing_25d') self.odm_georeferencing = io.join_paths(self.root_path, 'odm_georeferencing') self.odm_25dgeoreferencing = io.join_paths(self.root_path, 'odm_25dgeoreferencing') self.odm_orthophoto = io.join_paths(self.root_path, 'odm_orthophoto') self.odm_pdal = io.join_paths(self.root_path, 'pdal') # important files paths # benchmarking self.benchmarking = io.join_paths(self.root_path, 'benchmark.txt') # opensfm self.opensfm_tracks = io.join_paths(self.opensfm, 'tracks.csv') self.opensfm_bundle = io.join_paths(self.opensfm, 'bundle_r000.out') self.opensfm_bundle_list = io.join_paths(self.opensfm, 'list_r000.out') self.opensfm_image_list = io.join_paths(self.opensfm, 'image_list.txt') self.opensfm_reconstruction = io.join_paths(self.opensfm, 'reconstruction.json') self.opensfm_reconstruction_meshed = io.join_paths(self.opensfm, 'reconstruction.meshed.json') self.opensfm_reconstruction_nvm = io.join_paths(self.opensfm, 'reconstruction.nvm') self.opensfm_model = io.join_paths(self.opensfm, 'depthmaps/merged.ply') # pmvs self.pmvs_rec_path = io.join_paths(self.pmvs, 'recon0') self.pmvs_bundle = io.join_paths(self.pmvs_rec_path, 'bundle.rd.out') self.pmvs_visdat = io.join_paths(self.pmvs_rec_path, 'vis.dat') self.pmvs_options = io.join_paths(self.pmvs_rec_path, 'pmvs_options.txt') self.pmvs_model = io.join_paths(self.pmvs_rec_path, 'models/option-0000.ply') # odm_meshing self.odm_mesh = io.join_paths(self.odm_meshing, 'odm_mesh.ply') self.odm_meshing_log = io.join_paths(self.odm_meshing, 'odm_meshing_log.txt') self.odm_25dmesh = io.join_paths(self.odm_meshing, 'odm_25dmesh.ply') self.odm_25dmeshing_log = io.join_paths(self.odm_meshing, 'odm_25dmeshing_log.txt') # texturing self.odm_texturing_undistorted_image_path = io.join_paths( self.odm_texturing, 'undistorted') self.odm_textured_model_obj = 'odm_textured_model.obj' self.odm_textured_model_mtl = 'odm_textured_model.mtl' # Log is only used by old odm_texturing self.odm_texuring_log = 'odm_texturing_log.txt' # odm_georeferencing self.odm_georeferencing_latlon = io.join_paths( self.odm_georeferencing, 'latlon.txt') self.odm_georeferencing_coords = io.join_paths( self.odm_georeferencing, 'coords.txt') self.odm_georeferencing_gcp = io.join_paths( self.odm_georeferencing, 'gcp_list.txt') self.odm_georeferencing_utm_log = io.join_paths( self.odm_georeferencing, 'odm_georeferencing_utm_log.txt') self.odm_georeferencing_log = 'odm_georeferencing_log.txt' self.odm_georeferencing_transform_file = 'odm_georeferencing_transform.txt' self.odm_georeferencing_model_txt_geo = 'odm_georeferencing_model_geo.txt' self.odm_georeferencing_model_ply_geo = 'odm_georeferenced_model.ply' self.odm_georeferencing_model_obj_geo = 'odm_textured_model_geo.obj' self.odm_georeferencing_xyz_file = io.join_paths( self.odm_georeferencing, 'odm_georeferenced_model.csv') self.odm_georeferencing_las_json = io.join_paths( self.odm_georeferencing, 'las.json') self.odm_georeferencing_model_las = io.join_paths( self.odm_georeferencing, 'odm_georeferenced_model.las') self.odm_georeferencing_dem = io.join_paths( self.odm_georeferencing, 'odm_georeferencing_model_dem.tif') # odm_orthophoto self.odm_orthophoto_file = io.join_paths(self.odm_orthophoto, 'odm_orthophoto.png') self.odm_orthophoto_tif = io.join_paths(self.odm_orthophoto, 'odm_orthophoto.tif') self.odm_orthophoto_corners = io.join_paths(self.odm_orthophoto, 'odm_orthophoto_corners.txt') self.odm_orthophoto_log = io.join_paths(self.odm_orthophoto, 'odm_orthophoto_log.txt') self.odm_orthophoto_tif_log = io.join_paths(self.odm_orthophoto, 'gdal_translate_log.txt') self.odm_orthophoto_gdaladdo_log = io.join_paths(self.odm_orthophoto, 'gdaladdo_log.txt') def path(self, **args): return io.join_paths(self.root_path, **args)