Dense dsm step by step from sensor images¶
This notebook run step by step the dense dsm pipeline, from sensor images inputs to terrain DSM for one pair.
Imports¶
In [1]:
# Notebook local imports
import os
import math
###
# Silent OTB info logs
os.environ['OTB_LOGGER_LEVEL']='WARNING'
import warnings
# Filter warnings
warnings.filterwarnings("ignore",category=UserWarning)
# import pretty print
import pprint
pp = pprint.PrettyPrinter(indent=4)
# import external function for notebook
from notebook_helpers import get_full_data, show_data, save_data, get_dir_path, set_up_demo_inputs
In [2]:
# CARS imports
# Applications
from cars.applications.application import Application
from cars.applications.grid_generation import grid_correction
from cars.applications.sparse_matching import sparse_matching_tools
# Pipelines
import cars.pipelines.sensor_to_dense_dsm.sensor_dense_dsm_constants as sens_cst
from cars.pipelines.sensor_to_dense_dsm import sensors_inputs
from cars.pipelines.sensor_to_dense_dsm import dsm_output
# Conf, core, orchestrator
from cars.core import cars_logging
from cars.core import inputs, preprocessing
from cars.core.utils import safe_makedirs
from cars.orchestrator import orchestrator
from cars.core.utils import make_relative_path_absolute
In [3]:
# Show CARS version
from cars import __version__
print("CARS version used : {}".format(__version__))
CARS version used : 0.7.0
Inputs/Outputs/Applications Init¶
Define outputs¶
In [4]:
# Modify with your own output path if needed
output_dir = os.path.join(get_dir_path(), "output_notebook")
print(output_dir)
/home/vboxuser/Development/discover-cnes-3d-tools/cars/output_notebook
Define inputs¶
In [5]:
# By default, the notebook use data_gizeh_small.tar.bz2, data_gizeh is available also (performance dependent).
# For you own data: Modify input_dir_path and modify all images, geometric models and color file names below
input_dir_path = set_up_demo_inputs("data_gizeh_small")
inputs_conf = {
"sensors": {
"left": {
"image": os.path.join(input_dir_path, "img1.tif"),
"geomodel": os.path.join(input_dir_path, "img1.geom"),
"color": os.path.join(input_dir_path, "color1.tif"),
"no_data": 0,
},
"right": {
"image": os.path.join(input_dir_path, "img2.tif"),
"geomodel": os.path.join(input_dir_path, "img2.geom"),
"no_data": 0,
},
},
"pairing": [["left", "right"]],
"initial_elevation": os.path.join(input_dir_path, "srtm_dir", "N29E031_KHEOPS.tif")
}
inputs = sensors_inputs.sensors_check_inputs(inputs_conf)
# pp.pprint(inputs)
{ 'check_inputs': False,
'default_alt': 0,
'epipolar_a_priori': {},
'epsg': None,
'geoid': '/home/vboxuser/Development/cars/cars/pipelines/sensor_to_dense_dsm/../../conf/geoid/egm96.grd',
'initial_elevation': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/srtm_dir/N29E031_KHEOPS.tif',
'pairing': [['left', 'right']],
'roi': None,
'sensors': { 'left': { 'classification': None,
'color': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/color1.tif',
'geomodel': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/img1.geom',
'geomodel_filters': None,
'geomodel_type': 'RPC',
'image': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/img1.tif',
'mask': None,
'no_data': 0},
'right': { 'classification': None,
'color': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/img2.tif',
'geomodel': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/img2.geom',
'geomodel_filters': None,
'geomodel_type': 'RPC',
'image': '/home/vboxuser/Development/discover-cnes-3d-tools/cars/data_gizeh_small/img2.tif',
'mask': None,
'no_data': 0}},
'use_epipolar_a_priori': False}
Applications Init¶
GridGeneration¶
This application generates epipolar grids corresponding to sensor pair
In [6]:
epipolar_grid_generation_application = Application("grid_generation")
Resampling¶
This application generates epipolar images from epipolar grids
In [7]:
resampling_application = Application("resampling")
SparseMatching¶
This application generates sparse matches of stereo images pairs
In [8]:
sparse_matching_application = Application("sparse_matching")
DenseMatching¶
This application generates dense matches of stereo images pairs
In [9]:
dense_matching_application = Application("dense_matching")
Show used application configuration¶
In [10]:
# Example with dense matching application
# dense_matching_application.print_config()
{ 'epipolar_tile_margin_in_percent': 60,
'generate_performance_map': False,
'loader': 'pandora',
'loader_conf': OrderedDict([ ( 'input',
{ 'nodata_left': -9999,
'nodata_right': -9999}),
( 'pipeline',
{ 'cost_volume_confidence': { 'confidence_method': 'ambiguity',
'eta_max': 0.7,
'eta_step': 0.01,
'indicator': ''},
'disparity': { 'disparity_method': 'wta',
'invalid_disparity': nan},
'filter': { 'filter_method': 'median',
'filter_size': 3},
'matching_cost': { 'matching_cost_method': 'census',
'subpix': 1,
'window_size': 5},
'optimization': { 'min_cost_paths': False,
'optimization_method': 'sgm',
'overcounting': False,
'penalty': { 'P1': 8,
'P2': 32,
'p2_method': 'constant',
'penalty_method': 'sgm_penalty'},
'sgm_version': 'c++',
'use_confidence': False},
'refinement': { 'refinement_method': 'vfit'},
'right_disp_map': { 'method': 'accurate'},
'validation': { 'cross_checking_threshold': 1.0,
'validation_method': 'cross_checking'}})]),
'max_elevation_offset': None,
'max_epi_tile_size': 1500,
'method': 'census_sgm',
'min_elevation_offset': None,
'min_epi_tile_size': 300,
'perf_ambiguity_threshold': 0.6,
'perf_eta_max_ambiguity': 0.99,
'perf_eta_max_risk': 0.25,
'perf_eta_step': 0.04,
'save_disparity_map': False}
Triangulation¶
This application triangulates matches, in order to get each (X, Y, Z) point position
In [11]:
triangulation_application = Application("triangulation")
PointCloudFusion¶
This application performs the fusion of epipolar points from pairs to a terrain point cloud
In [12]:
pc_fusion_application = Application("point_cloud_fusion")
PointCloudOutliersRemoving : small components¶
This application removes outliers points. The method used is the "small components removing"
In [13]:
conf_outlier_removing_small_components = {"method": "small_components", "activated": True}
pc_outlier_removing_small_comp_application = Application("point_cloud_outliers_removing", cfg=conf_outlier_removing_small_components)
PointCloudOutliersRemoving : statistical¶
This application removes outliers points. The method used is the "statistical removing"
In [14]:
conf_outlier_removing_small_statistical = {"method": "statistical", "activated": True}
pc_outlier_removing_stats_application = Application("point_cloud_outliers_removing", cfg=conf_outlier_removing_small_statistical)
PointCloudRasterization¶
This application performs the rasterization of a terrain point cloint.
In [15]:
conf_rasterization = {
"method": "simple_gaussian",
"dsm_radius": 3,
"sigma": 0.3
}
rasterization_application = Application("point_cloud_rasterization", cfg=conf_rasterization)
Create orchestrator¶
In [16]:
# Use sequential mode in notebook
orchestrator_conf = {"mode": "sequential"}
cars_orchestrator = orchestrator.Orchestrator(orchestrator_conf=orchestrator_conf, out_dir=output_dir)
Run pipeline step by step from sensors to DSM¶
Sensors images generation¶
From input configuration "inputs" seen before
In [17]:
_, sensor_image_left, sensor_image_right = sensors_inputs.generate_inputs(inputs)[0]
Grid Generation : epipolar grid generation¶
In [18]:
grid_left, grid_right = epipolar_grid_generation_application.run(
sensor_image_left,
sensor_image_right,
orchestrator=cars_orchestrator,
srtm_dir=inputs[sens_cst.INITIAL_ELEVATION],
default_alt=inputs[sens_cst.DEFAULT_ALT],
geoid_path=inputs[sens_cst.GEOID],
)
Resampling : epipolar images generation¶
In [19]:
epipolar_image_left, epipolar_image_right = resampling_application.run(
sensor_image_left,
sensor_image_right,
grid_left,
grid_right,
orchestrator=cars_orchestrator,
margins=sparse_matching_application.get_margins()
)
Show epipolar image¶
In [20]:
data_image_left = get_full_data(epipolar_image_left, "im")
show_data(data_image_left, mode="image")
Sparse matching: compute sifts¶
In [21]:
epipolar_matches_left, _ = sparse_matching_application.run(
epipolar_image_left,
epipolar_image_right,
grid_left.attributes["disp_to_alt_ratio"],
orchestrator=cars_orchestrator
)
Grid correction: correct epipolar grids from sparse matches¶
Find correction to apply, and generate new right epipolar grid
Filter matches¶
In [22]:
matches_array = sparse_matching_application.filter_matches(epipolar_matches_left, orchestrator=cars_orchestrator)
Estimate grid correction¶
In [23]:
grid_correction_coef, corrected_matches_array, _, _, _ = grid_correction.estimate_right_grid_correction(matches_array, grid_right)
Correct right grid¶
In [24]:
corrected_grid_right = grid_correction.correct_grid(grid_right, grid_correction_coef)
Estimate disp min and disp max from sparse matches¶
In [25]:
dmin, dmax = sparse_matching_tools.compute_disp_min_disp_max(
sensor_image_left,
sensor_image_right,
grid_left,
corrected_grid_right,
grid_right,
corrected_matches_array,
orchestrator=cars_orchestrator,
disp_margin=(
sparse_matching_application.get_disparity_margin()
),
disp_to_alt_ratio=grid_left.attributes["disp_to_alt_ratio"],
geometry_loader=triangulation_application.get_geometry_loader(),
srtm_dir=inputs[sens_cst.INITIAL_ELEVATION],
default_alt=inputs[sens_cst.DEFAULT_ALT],
)
Compute margins used in dense matching, with corresponding disparity min and max¶
In [26]:
dense_matching_margins, disp_min, disp_max = dense_matching_application.get_margins(
grid_left, disp_min=dmin, disp_max=dmax)
Resampling: generate epipolar images with corrected grids and new margins¶
In [27]:
new_epipolar_image_left, new_epipolar_image_right = resampling_application.run(
sensor_image_left,
sensor_image_right,
grid_left,
corrected_grid_right,
orchestrator=cars_orchestrator,
margins=dense_matching_margins,
optimum_tile_size=(
dense_matching_application.get_optimal_tile_size(
disp_min,
disp_max,
cars_orchestrator.cluster.checked_conf_cluster[
"max_ram_per_worker"
],
)
),
add_color=True,
)
Dense Matching: compute disparities with pandora¶
In [28]:
epipolar_disparity_map = dense_matching_application.run(
new_epipolar_image_left,
new_epipolar_image_right,
orchestrator=cars_orchestrator,
disp_min=disp_min,
disp_max=disp_max,
)
Show full disparity map¶
In [29]:
data_disparity = get_full_data(epipolar_disparity_map, "disp")
show_data(data_disparity)
Compute epsg
In [30]:
epsg = preprocessing.compute_epsg(
sensor_image_left,
sensor_image_right,
grid_left,
corrected_grid_right,
triangulation_application.get_geometry_loader(),
orchestrator=cars_orchestrator,
srtm_dir=inputs[sens_cst.INITIAL_ELEVATION],
default_alt=inputs[sens_cst.DEFAULT_ALT],
disp_min=disp_min,
disp_max=disp_max
)
Triangulation : triangulate matches¶
In [31]:
epipolar_points_cloud = triangulation_application.run(
sensor_image_left,
sensor_image_right,
new_epipolar_image_left,
grid_left,
corrected_grid_right,
epipolar_disparity_map,
epsg,
orchestrator=cars_orchestrator,
uncorrected_grid_right=grid_right,
geoid_path=inputs[sens_cst.GEOID],
disp_min=disp_min,
disp_max=disp_max,
)
Compute terrain bounding box¶
In [32]:
current_terrain_roi_bbox = preprocessing.compute_terrain_bbox(
inputs[sens_cst.INITIAL_ELEVATION],
inputs[sens_cst.DEFAULT_ALT],
inputs[sens_cst.GEOID],
sensor_image_left,
sensor_image_right,
new_epipolar_image_left,
grid_left,
corrected_grid_right,
epsg,
triangulation_application.get_geometry_loader(),
resolution=rasterization_application.get_resolution(),
disp_min=disp_min,
disp_max=disp_max,
orchestrator=cars_orchestrator
)
terrain_bounds, optimal_terrain_tile_width = preprocessing.compute_terrain_bounds(
[current_terrain_roi_bbox],
resolution=rasterization_application.get_resolution()
)
Transform point cloud to terrain point cloud¶
In [33]:
merged_points_clouds = pc_fusion_application.run(
[epipolar_points_cloud],
terrain_bounds,
epsg,
orchestrator=cars_orchestrator,
margins=rasterization_application.get_margins(),
optimal_terrain_tile_width=optimal_terrain_tile_width
)
Point Cloud Outlier Removing : remove points with small components removing method¶
In [34]:
filtered_sc_merged_points_clouds = pc_outlier_removing_small_comp_application.run(
merged_points_clouds,
orchestrator=cars_orchestrator,
)
Point Cloud Outlier Removing: remove points with statistical removing method¶
In [35]:
filtered_stats_merged_points_clouds = pc_outlier_removing_stats_application.run(
filtered_sc_merged_points_clouds,
orchestrator=cars_orchestrator,
)
Rasterization : rasterize point cloud¶
In [36]:
dsm = rasterization_application.run(
filtered_stats_merged_points_clouds,
epsg,
orchestrator=cars_orchestrator
)
Show DSM¶
In [37]:
data_dsm = get_full_data(dsm, "hgt")
show_data(data_dsm, mode="dsm")
Show ortho image¶
In [38]:
data_ortho = get_full_data(dsm, "img")[..., 0:3]
show_data(data_ortho, mode='image')
Save DSM¶
In [39]:
save_data(dsm, os.path.join(output_dir, "dsm.tif"), "hgt")
Dense dsm step by step from sensor images This notebook run step by step the dense dsm pipeline, from sensor images inputs to terrain DSM for one pair.