cars 1.0.0rc3__cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

cars/applications/rasterization/simple_gaussian_app.py

@@ -0,0 +1,1152 @@
+#!/usr/bin/env python
+# coding: utf8
+#
+# Copyright (c) 2020 Centre National d'Etudes Spatiales (CNES).
+#
+# This file is part of CARS
+# (see https://github.com/CNES/cars).
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# pylint: disable=too-many-lines
+
+"""
+this module contains the dense_matching application class.
+"""
+
+import collections
+
+# Third party imports
+import copy
+
+# Standard imports
+import logging
+import os
+from typing import List
+
+import numpy as np
+import rasterio as rio
+import xarray
+from affine import Affine
+from json_checker import Checker, Or
+
+import cars.orchestrator.orchestrator as ocht
+from cars.applications import application_constants
+
+# CARS imports
+from cars.applications.rasterization import (
+    rasterization_algo,
+)
+from cars.applications.rasterization import (
+    rasterization_constants as raster_cst,
+)
+from cars.applications.rasterization import (
+    rasterization_wrappers,
+)
+from cars.applications.rasterization.abstract_pc_rasterization_app import (
+    PointCloudRasterization,
+)
+from cars.applications.triangulation import pc_transform
+from cars.core import constants as cst
+from cars.core import projection, tiling
+from cars.core.utils import safe_makedirs
+from cars.data_structures import cars_dataset
+
+# R0903  temporary disabled for error "Too few public methods"
+# œgoing to be corrected by adding new methods as check_conf
+
+
+class SimpleGaussian(
+    PointCloudRasterization, short_name="simple_gaussian"
+):  # pylint: disable=R0903
+    """
+    PointCloudRasterisation
+    """
+
+    # pylint: disable=too-many-instance-attributes
+
+    def __init__(self, conf=None):
+        """
+        Init function of PointCloudRasterisation
+
+        :param conf: configuration for rasterization
+        :return: a application_to_use object
+        """
+
+        super().__init__(conf=conf)
+
+        # check conf
+
+        # get rasterization parameter
+        self.used_method = self.used_config["method"]
+        self.dsm_radius = self.used_config["dsm_radius"]
+        self.sigma = self.used_config["sigma"]
+        self.grid_points_division_factor = self.used_config[
+            "grid_points_division_factor"
+        ]
+        # get nodata values
+        self.dsm_no_data = self.used_config["dsm_no_data"]
+        self.texture_no_data = self.used_config["texture_no_data"]
+        self.color_dtype = self.used_config["texture_dtype"]
+        self.msk_no_data = self.used_config["msk_no_data"]
+
+        # Init orchestrator
+        self.orchestrator = None
+
+    def check_conf(self, conf):
+        """
+        Check configuration
+
+        :param conf: configuration to check
+        :type conf: dict
+
+        :return: overloaded configuration
+        :rtype: dict
+
+        """
+
+        # init conf
+        if conf is not None:
+            overloaded_conf = conf.copy()
+        else:
+            conf = {}
+            overloaded_conf = {}
+
+        # Overload conf
+
+        # get rasterization parameter
+        overloaded_conf["method"] = conf.get("method", "simple_gaussian")
+        overloaded_conf["dsm_radius"] = conf.get("dsm_radius", 1)
+        overloaded_conf["sigma"] = conf.get("sigma", None)
+        overloaded_conf["grid_points_division_factor"] = conf.get(
+            "grid_points_division_factor", None
+        )
+
+        # get nodata values
+        overloaded_conf["dsm_no_data"] = conf.get("dsm_no_data", -32768)
+        overloaded_conf["texture_no_data"] = conf.get("texture_no_data", None)
+        overloaded_conf["texture_dtype"] = conf.get("texture_dtype", None)
+        overloaded_conf["msk_no_data"] = conf.get("msk_no_data", 255)
+
+        overloaded_conf["save_intermediate_data"] = conf.get(
+            "save_intermediate_data", False
+        )
+
+        rasterization_schema = {
+            "method": str,
+            "dsm_radius": Or(float, int),
+            "sigma": Or(float, None),
+            "grid_points_division_factor": Or(None, int),
+            "dsm_no_data": int,
+            "msk_no_data": int,
+            "texture_no_data": Or(None, int),
+            "texture_dtype": Or(None, str),
+            "save_intermediate_data": bool,
+        }
+
+        # Check conf
+        checker = Checker(rasterization_schema)
+        checker.validate(overloaded_conf)
+
+        return overloaded_conf
+
+    def get_margins(self, resolution):
+        """
+        Get the margin to use for terrain tiles
+
+        :param resolution: resolution of raster data (in target CRS unit)
+        :type resolution: float
+
+        :return: margin in meters or degrees
+        """
+
+        margins = self.dsm_radius * resolution
+        return margins
+
+    def get_optimal_tile_size(
+        self,
+        max_ram_per_worker,
+        superposing_point_clouds=1,
+        point_cloud_resolution=0.5,
+    ):
+        """
+        Get the optimal tile size to use, depending on memory available
+
+        :param max_ram_per_worker: maximum ram available
+        :type max_ram_per_worker: int
+        :param superposing_point_clouds: number of point clouds superposing
+        :type superposing_point_clouds: int
+        :param point_cloud_resolution: resolution of point cloud
+        :type point_cloud_resolution: float
+
+        :return: optimal tile size in meter
+        :rtype: float
+
+        """
+
+        tot = 7000 * superposing_point_clouds / point_cloud_resolution
+
+        import_ = 200  # MiB
+        tile_size = int(
+            np.sqrt(float(((max_ram_per_worker - import_) * 2**23)) / tot)
+        )
+
+        logging.info(
+            "Estimated optimal tile size for rasterization: {} meters".format(
+                tile_size
+            )
+        )
+        return tile_size
+
+    # pylint: disable=too-many-positional-arguments
+    def run(  # noqa: C901 function is too complex
+        self,
+        point_clouds,
+        epsg,
+        output_crs,
+        resolution,
+        orchestrator=None,
+        dsm_file_name=None,
+        weights_file_name=None,
+        color_file_name=None,
+        classif_file_name=None,
+        performance_map_file_name=None,
+        ambiguity_file_name=None,
+        contributing_pair_file_name=None,
+        filling_file_name=None,
+        color_dtype=None,
+        dump_dir=None,
+        performance_map_classes=None,
+        phasing=None,
+    ):
+        """
+        Run PointCloudRasterisation application.
+
+        Creates a CarsDataset filled with dsm tiles.
+
+        :param point_clouds: merged point cloud or list of array point clouds
+
+            . CarsDataset contains:
+
+            - Z x W Delayed tiles. \
+                Each tile will be a future pandas DataFrame containing:
+
+                - data with keys  "x", "y", "z", "corr_msk" \
+                    optional: "texture", "mask", "data_valid", "z_inf", "z_sup"\
+                      "coord_epi_geom_i", "coord_epi_geom_j", "idx_im_epi"
+                - attrs with keys "epsg", "ysize", "xsize", "xstart", "ystart"
+
+             - attributes containing "bounds", "ysize", "xsize", "epsg"
+
+             OR
+
+
+             Tuple(list of CarsDataset Arrays, bounds). With list of point
+                clouds:
+                list of CarsDataset of type array, with:
+                - data with keys x", "y", "z", "corr_msk", "z_inf", "z_sup"\
+                    optional: "texture", "mask", "data_valid",\
+                      "coord_epi_geom_i", "coord_epi_geom_j", "idx_im_epi"
+
+        :type point_clouds: CarsDataset filled with pandas.DataFrame
+        :param epsg: epsg of raster data
+        :type epsg: str
+        :param output_crs: output_crs of raster data
+        :type output_crs: str
+        :param resolution: resolution of raster data (in target CRS unit)
+        :type resolution: float
+        :param orchestrator: orchestrator used
+        :param dsm_file_name: path of dsm
+        :type dsm_file_name: str
+        :param weights_file_name: path of dsm weights
+        :type weights_file_name: str
+        :param color_file_name: path of color
+        :type color_file_name: str
+        :param classif_file_name: path of classification
+        :type classif_file_name: str
+        :param performance_map_file_name: path of performance map file
+        :type performance_map_file_name: str
+        :param ambiguity_file_name: path of ambiguity file
+        :type ambiguity_file_name: str
+        :param contributing_pair_file_name: path of contributing pair file
+        :type contributing_pair_file_name: str
+        :param filling_file_name: path of filling file
+        :type filling_file_name: str
+        :param color_dtype: output color image type
+        :type color_dtype: str (numpy type)
+        :param dump_dir: directory used for outputs with no associated filename
+        :type dump_dir: str
+        :param performance_map_classes: list for step defining border of class
+        :type performance_map_classes: list or None
+        :param phasing: if activated, we phase the dsm on this point
+        :type phasing: dict
+
+        :return: raster DSM. CarsDataset contains:
+
+            - Z x W Delayed tiles. \
+                Each tile will be a future xarray Dataset containing:
+
+                - data : with keys : "hgt", "img", "raster_msk",optional : \
+                  "n_pts", "pts_in_cell", "hgt_mean", "hgt_stdev",\
+                  "hgt_inf", "hgt_sup"
+                - attrs with keys: "epsg"
+            - attributes containing: None
+
+        :rtype : CarsDataset filled with xr.Dataset
+        """
+        # only the saved layers will be saved
+        list_computed_layers = []
+
+        # Default orchestrator
+        if orchestrator is None:
+            # Create default sequential orchestrator for current application
+            # be awere, no out_json will be shared between orchestrators
+            # No files saved
+            self.orchestrator = ocht.Orchestrator(
+                orchestrator_conf={"mode": "sequential"}
+            )
+        else:
+            self.orchestrator = orchestrator
+
+        # Get if color and stats are saved
+        save_intermediate_data = self.used_config["save_intermediate_data"]
+
+        keep_dir = (
+            len(
+                list(
+                    filter(
+                        lambda x: x is not None,
+                        [
+                            weights_file_name,
+                            color_file_name,
+                            classif_file_name,
+                            performance_map_file_name,
+                            ambiguity_file_name,
+                            contributing_pair_file_name,
+                            filling_file_name,
+                        ],
+                    )
+                )
+            )
+            > 0
+        )
+
+        if not self.color_dtype:
+            self.color_dtype = color_dtype
+
+        if self.texture_no_data is None:
+            if self.color_dtype is not None:
+                if "float" in self.color_dtype:
+                    self.texture_no_data = np.finfo(self.color_dtype).max
+                else:
+                    self.texture_no_data = np.iinfo(self.color_dtype).max
+            else:
+                self.texture_no_data = np.finfo("float32").max
+
+        # Setup dump directory
+        if dump_dir is not None:
+            out_dump_dir = dump_dir
+            safe_makedirs(dump_dir)
+            if not save_intermediate_data and not keep_dir:
+                self.orchestrator.add_to_clean(dump_dir)
+        else:
+            out_dump_dir = self.orchestrator.out_dir
+
+        # Check if input data is supported
+        if not (
+            isinstance(point_clouds, tuple)
+            and isinstance(point_clouds[0][0], cars_dataset.CarsDataset)
+            and point_clouds[0][0].dataset_type == "arrays"
+        ):
+            message = (
+                "PointCloudRasterization application doesn't support "
+                "this input data "
+                "format : type : {}".format(type(point_clouds))
+            )
+            logging.error(message)
+            raise RuntimeError(message)
+
+        # Create CarsDataset
+        terrain_raster = cars_dataset.CarsDataset(
+            "arrays", name="rasterization"
+        )
+
+        paths_data = {}
+
+        bounds = point_clouds[1]
+        # tiling grid: all tiles from sources -> not replaceable.
+        # CarsDataset is only used for processing
+        nb_tiles = 0
+        for point_cld in point_clouds[0]:
+            nb_tiles += point_cld.shape[0] * point_cld.shape[1]
+        terrain_raster.tiling_grid = tiling.generate_tiling_grid(
+            0, 0, 1, nb_tiles, 1, 1
+        )
+
+        terrain_raster.generate_none_tiles()
+
+        if phasing is not None:
+            res = resolution
+            x_phase = phasing["point"][0]
+            y_phase = phasing["point"][1]
+
+            for index, value in enumerate(bounds):
+                if index in (0, 2):
+                    bounds[index] = rasterization_wrappers.phased_dsm(
+                        value, x_phase, res
+                    )
+                else:
+                    bounds[index] = rasterization_wrappers.phased_dsm(
+                        value, y_phase, res
+                    )
+
+        # Derive output image files parameters to pass to rasterio
+        _, _, xsize, ysize = tiling.roi_to_start_and_size(bounds, resolution)
+        logging.info("DSM output image size: {}x{} pixels".format(xsize, ysize))
+
+        try:
+            if isinstance(point_clouds, tuple):
+                source_pc_names = point_clouds[0][0].attributes[
+                    "source_pc_names"
+                ]
+            else:
+                source_pc_names = point_clouds.attributes["source_pc_names"]
+        except KeyError:
+            source_pc_names = None
+
+        # Save objects
+
+        if dsm_file_name is not None:
+            safe_makedirs(os.path.dirname(dsm_file_name))
+
+        out_dsm_file_name = dsm_file_name
+
+        if out_dsm_file_name is not None:
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_ALT: os.path.basename(out_dsm_file_name)
+                    }
+                }
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_dsm_file_name = os.path.join(out_dump_dir, "dsm.tif")
+        if out_dsm_file_name is not None:
+            list_computed_layers += ["dsm"]
+            self.orchestrator.add_to_save_lists(
+                out_dsm_file_name,
+                cst.RASTER_HGT,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm",
+            )
+            paths_data[cst.RASTER_HGT] = out_dsm_file_name
+
+        out_weights_file_name = weights_file_name
+        if out_weights_file_name is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_WEIGHTS: os.path.basename(
+                            out_weights_file_name
+                        )
+                    }
+                }
+            )
+        else:
+            # Always write weights.tif, but dump_dir is in the orchestrator
+            # clean list if save_intermediate_data is not activated
+            out_weights_file_name = os.path.join(out_dump_dir, "weights.tif")
+        if out_weights_file_name is not None:
+            list_computed_layers += ["weights"]
+            self.orchestrator.add_to_save_lists(
+                out_weights_file_name,
+                cst.RASTER_WEIGHTS_SUM,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=0,
+                cars_ds_name="dsm_weights",
+            )
+            paths_data[cst.RASTER_WEIGHTS_SUM] = out_weights_file_name
+
+        out_clr_file_name = color_file_name
+        if out_clr_file_name is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_COLOR: os.path.basename(out_clr_file_name)
+                    }
+                }
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_clr_file_name = os.path.join(out_dump_dir, "image.tif")
+        if out_clr_file_name is not None:
+            list_computed_layers += ["texture"]
+
+            self.orchestrator.add_to_save_lists(
+                out_clr_file_name,
+                cst.RASTER_COLOR_IMG,
+                terrain_raster,
+                dtype=self.color_dtype,
+                nodata=self.texture_no_data,
+                cars_ds_name="texture",
+            )
+            paths_data[cst.RASTER_COLOR_IMG] = out_clr_file_name
+
+        out_classif_file_name = classif_file_name
+        if out_classif_file_name is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_CLASSIFICATION: os.path.basename(
+                            out_classif_file_name
+                        )
+                    }
+                }
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_classif_file_name = os.path.join(
+                out_dump_dir, "classification.tif"
+            )
+
+        if out_classif_file_name is not None:
+            list_computed_layers += ["classif"]
+            self.orchestrator.add_to_save_lists(
+                out_classif_file_name,
+                cst.RASTER_CLASSIF,
+                terrain_raster,
+                dtype=np.uint8,
+                nodata=self.msk_no_data,
+                cars_ds_name="dsm_classif",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_CLASSIF] = out_classif_file_name
+
+        out_performance_map = performance_map_file_name
+
+        # save raw as performance map if performance_map_classes is None
+        # save classified performance map if performance_map_classes is not None
+        out_performance_map_raw = None
+        if out_performance_map is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_PERFORMANCE_MAP: os.path.basename(
+                            out_performance_map
+                        )
+                    }
+                }
+            )
+            if performance_map_classes is False:
+                # No classes, we return raw data
+                out_performance_map_raw = out_performance_map
+                out_performance_map = None
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_performance_map = os.path.join(
+                out_dump_dir, "performance_map.tif"
+            )
+            out_performance_map_raw = os.path.join(
+                out_dump_dir, "performance_map_raw.tif"
+            )
+
+        if out_performance_map_raw is not None:
+            list_computed_layers += ["performance_map_raw"]
+            self.orchestrator.add_to_save_lists(
+                out_performance_map_raw,
+                cst.RASTER_PERFORMANCE_MAP_RAW,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.msk_no_data,
+                cars_ds_name="performance_map_raw",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_PERFORMANCE_MAP_RAW] = out_performance_map_raw
+
+        if out_performance_map is not None:
+            # classified performance map exists
+            list_computed_layers += ["performance_map"]
+            self.orchestrator.add_to_save_lists(
+                out_performance_map,
+                cst.RASTER_PERFORMANCE_MAP,
+                terrain_raster,
+                dtype=np.uint8,
+                nodata=self.msk_no_data,
+                cars_ds_name="performance_map",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_PERFORMANCE_MAP] = out_performance_map
+
+        out_ambiguity = ambiguity_file_name
+        if out_ambiguity is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_AMBIGUITY: os.path.basename(out_ambiguity)
+                    }
+                }
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_ambiguity = os.path.join(out_dump_dir, "ambiguity.tif")
+        if out_ambiguity:
+            list_computed_layers += ["ambiguity"]
+            self.orchestrator.add_to_save_lists(
+                out_ambiguity,
+                cst.RASTER_AMBIGUITY,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.msk_no_data,
+                cars_ds_name="ambiguity",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_AMBIGUITY] = out_ambiguity
+
+        out_source_pc = contributing_pair_file_name
+        if out_source_pc is not None:
+            # add contributing pair filename to index
+            self.orchestrator.update_index(
+                {
+                    "dsm": {
+                        cst.INDEX_DSM_CONTRIBUTING_PAIR: os.path.basename(
+                            out_source_pc
+                        )
+                    }
+                }
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_source_pc = os.path.join(out_dump_dir, "contributing_pair.tif")
+        if out_source_pc:
+            list_computed_layers += ["source_pc"]
+            self.orchestrator.add_to_save_lists(
+                out_source_pc,
+                cst.RASTER_SOURCE_PC,
+                terrain_raster,
+                dtype=np.uint8,
+                nodata=self.msk_no_data,
+                cars_ds_name="contributing_pair",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_SOURCE_PC] = out_source_pc
+
+        out_filling = filling_file_name
+        if out_filling is not None:
+            # add filling filename to index
+            self.orchestrator.update_index(
+                {"dsm": {cst.INDEX_DSM_FILLING: os.path.basename(out_filling)}}
+            )
+        elif save_intermediate_data:
+            # File is not part of the official product, write it in dump_dir
+            out_filling = os.path.join(out_dump_dir, "filling.tif")
+        if out_filling:
+            list_computed_layers += ["filling"]
+            self.orchestrator.add_to_save_lists(
+                out_filling,
+                cst.RASTER_FILLING,
+                terrain_raster,
+                dtype=np.uint8,
+                nodata=self.msk_no_data,
+                cars_ds_name="filling",
+                optional_data=True,
+            )
+            paths_data[cst.RASTER_FILLING] = out_filling
+
+        # TODO Check that intervals indeed exist!
+        if save_intermediate_data:
+            list_computed_layers += [cst.POINT_CLOUD_LAYER_SUP_OR_INF_ROOT]
+            out_dsm_inf_file_name = os.path.join(out_dump_dir, "dsm_inf.tif")
+            self.orchestrator.add_to_save_lists(
+                out_dsm_inf_file_name,
+                cst.RASTER_HGT_INF,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_inf",
+                optional_data=True,
+            )
+            out_dsm_sup_file_name = os.path.join(out_dump_dir, "dsm_sup.tif")
+            self.orchestrator.add_to_save_lists(
+                out_dsm_sup_file_name,
+                cst.RASTER_HGT_SUP,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_sup",
+                optional_data=True,
+            )
+
+            out_dsm_mean_file_name = os.path.join(out_dump_dir, "dsm_mean.tif")
+            out_dsm_std_file_name = os.path.join(out_dump_dir, "dsm_std.tif")
+            out_dsm_n_pts_file_name = os.path.join(
+                out_dump_dir, "dsm_n_pts.tif"
+            )
+            out_dsm_points_in_cell_file_name = os.path.join(
+                out_dump_dir, "dsm_pts_in_cell.tif"
+            )
+            self.orchestrator.add_to_save_lists(
+                out_dsm_mean_file_name,
+                cst.RASTER_HGT_MEAN,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_mean",
+            )
+            self.orchestrator.add_to_save_lists(
+                out_dsm_std_file_name,
+                cst.RASTER_HGT_STD_DEV,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_std",
+            )
+            self.orchestrator.add_to_save_lists(
+                out_dsm_n_pts_file_name,
+                cst.RASTER_NB_PTS,
+                terrain_raster,
+                dtype=np.uint16,
+                nodata=0,
+                cars_ds_name="dsm_n_pts",
+            )
+            self.orchestrator.add_to_save_lists(
+                out_dsm_points_in_cell_file_name,
+                cst.RASTER_NB_PTS_IN_CELL,
+                terrain_raster,
+                dtype=np.uint16,
+                nodata=0,
+                cars_ds_name="dsm_pts_in_cells",
+            )
+
+            out_dsm_inf_mean_file_name = os.path.join(
+                out_dump_dir, "dsm_inf_mean.tif"
+            )
+            out_dsm_inf_std_file_name = os.path.join(
+                out_dump_dir, "dsm_inf_std.tif"
+            )
+
+            self.orchestrator.add_to_save_lists(
+                out_dsm_inf_mean_file_name,
+                cst.RASTER_HGT_INF_MEAN,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_inf_mean",
+                optional_data=True,
+            )
+
+            self.orchestrator.add_to_save_lists(
+                out_dsm_inf_std_file_name,
+                cst.RASTER_HGT_INF_STD_DEV,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_inf_std",
+                optional_data=True,
+            )
+
+            out_dsm_sup_mean_file_name = os.path.join(
+                out_dump_dir, "dsm_sup_mean.tif"
+            )
+            out_dsm_sup_std_file_name = os.path.join(
+                out_dump_dir, "dsm_sup_std.tif"
+            )
+
+            self.orchestrator.add_to_save_lists(
+                out_dsm_sup_mean_file_name,
+                cst.RASTER_HGT_SUP_MEAN,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_sup_mean",
+                optional_data=True,
+            )
+            self.orchestrator.add_to_save_lists(
+                out_dsm_sup_std_file_name,
+                cst.RASTER_HGT_SUP_STD_DEV,
+                terrain_raster,
+                dtype=np.float32,
+                nodata=self.dsm_no_data,
+                cars_ds_name="dsm_sup_std",
+                optional_data=True,
+            )
+
+        # Get saving infos in order to save tiles when they are computed
+        [saving_info] = self.orchestrator.get_saving_infos([terrain_raster])
+
+        # Generate profile
+        geotransform = (
+            bounds[0],
+            resolution,
+            0.0,
+            bounds[3],
+            0.0,
+            -resolution,
+        )
+
+        transform = Affine.from_gdal(*geotransform)
+        raster_profile = collections.OrderedDict(
+            {
+                "height": ysize,
+                "width": xsize,
+                "driver": "GTiff",
+                "dtype": "float32",
+                "transform": transform,
+                "crs": output_crs.to_wkt(),
+                "tiled": True,
+            }
+        )
+
+        # Get number of tiles
+        logging.info(
+            "Number of tiles in cloud rasterization: "
+            "row: {} "
+            "col: {}".format(
+                terrain_raster.tiling_grid.shape[0],
+                terrain_raster.tiling_grid.shape[1],
+            )
+        )
+
+        # Add infos to orchestrator.out_json
+        updating_dict = {
+            application_constants.APPLICATION_TAG: {
+                raster_cst.RASTERIZATION_RUN_TAG: {
+                    raster_cst.EPSG_TAG: epsg,
+                    raster_cst.DSM_NO_DATA_TAG: float(self.dsm_no_data),
+                },
+            }
+        }
+        if self.texture_no_data is not None:
+            updating_dict[application_constants.APPLICATION_TAG][
+                raster_cst.RASTERIZATION_RUN_TAG
+            ][raster_cst.TEXTURE_NO_DATA_TAG] = float(self.texture_no_data)
+
+        self.orchestrator.update_out_info(updating_dict)
+
+        # Add attributrs
+        terrain_raster.attributes["paths"] = paths_data
+
+        # Add final function to apply
+        terrain_raster.final_function = raster_final_function
+        ind_tile = 0
+        for point_cloud in point_clouds[0]:
+            for row_pc in range(point_cloud.shape[0]):
+                for col_pc in range(point_cloud.shape[1]):
+                    # update saving infos  for potential replacement
+                    full_saving_info = ocht.update_saving_infos(
+                        saving_info, row=0, col=ind_tile
+                    )
+                    if point_cloud[row_pc, col_pc] is not None:
+                        # Delayed call to rasterization operations using all
+                        #  required point clouds
+                        terrain_raster[
+                            0, ind_tile
+                        ] = self.orchestrator.cluster.create_task(
+                            rasterization_wrapper
+                        )(
+                            point_cloud[row_pc, col_pc],
+                            resolution,
+                            epsg,
+                            raster_profile,
+                            window=None,
+                            terrain_region=None,
+                            terrain_full_roi=bounds,
+                            list_computed_layers=list_computed_layers,
+                            saving_info=full_saving_info,
+                            radius=self.dsm_radius,
+                            sigma=self.sigma,
+                            dsm_no_data=self.dsm_no_data,
+                            texture_no_data=self.texture_no_data,
+                            color_dtype=color_dtype,
+                            msk_no_data=self.msk_no_data,
+                            source_pc_names=source_pc_names,
+                            performance_map_classes=performance_map_classes,
+                        )
+                    ind_tile += 1
+
+        # Sort tiles according to rank TODO remove or implement it ?
+
+        return terrain_raster
+
+
+# pylint: disable=too-many-positional-arguments
+def rasterization_wrapper(  # noqa: C901
+    cloud,
+    resolution,
+    epsg,
+    profile,
+    window=None,
+    terrain_region=None,
+    terrain_full_roi=None,
+    list_computed_layers: List[str] = None,
+    saving_info=None,
+    sigma: float = None,
+    radius: int = 1,
+    dsm_no_data: int = np.nan,
+    texture_no_data: int = np.nan,
+    color_dtype: str = "float32",
+    msk_no_data: int = 255,
+    source_pc_names=None,
+    performance_map_classes=None,
+):
+    """
+    Wrapper for rasterization step :
+    - Convert a list of clouds to correct epsg
+    - Rasterize it with associated colors
+
+    if terrain_region is not provided: region is computed from point cloud,
+        with margin to use
+
+    :param cloud: combined cloud
+    :type cloud: pandas.DataFrame
+    :param terrain_region: terrain bounds
+    :param resolution: Produced DSM resolution (meter, degree [EPSG dependent])
+    :type resolution: float
+    :param  epsg_code: epsg code for the CRS of the output DSM
+    :type epsg_code: int
+    :param  window: Window considered
+    :type window: int
+    :param  margin: margin in pixel to use
+    :type margin: int
+    :param  profile: rasterio profile
+    :param list_computed_layers: list of computed output data
+    :type profile: dict
+    :param saving_info: information about CarsDataset ID.
+    :type saving_info: dict
+    :param sigma: sigma for gaussian interpolation.
+        (If None, set to resolution)
+    :param radius: Radius for hole filling.
+    :param dsm_no_data: no data value to use in the final raster
+    :param texture_no_data: no data value to use in the final colored raster
+    :param msk_no_data: no data value to use in the final mask image
+    :param source_pc_names: list of names of point cloud before merging :
+        name of sensors pair or name of point cloud file
+    :param performance_map_classes: list for step defining border of class
+    :type performance_map_classes: list or None
+
+    :return: digital surface model + projected colors
+    :rtype: xr.Dataset
+    """
+    # update attributes
+    attributes = copy.deepcopy(cloud.attrs)
+    attributes.update(attributes.get("attributes", {}))
+    if "attributes" in attributes:
+        del attributes["attributes"]
+    if "saving_info" in attributes:
+        del attributes["saving_info"]
+
+    # convert back to correct epsg
+    # If the point cloud is not in the right epsg referential, it is converted
+    if isinstance(cloud, xarray.Dataset):
+        # Transform Dataset to Dataframe
+        cloud, cloud_epsg = pc_transform.depth_map_dataset_to_dataframe(
+            cloud, epsg
+        )
+    elif cloud is None:
+        logging.warning("Input cloud is None")
+        return None
+    else:
+        cloud_epsg = attributes.get("epsg")
+
+    if "number_of_pc" not in attributes:
+        if source_pc_names is not None:
+            attributes["number_of_pc"] = len(source_pc_names)
+        else:
+            attributes["number_of_pc"] = None
+
+    # update attributes
+    cloud.attrs = {}
+    cars_dataset.fill_dataframe(cloud, attributes=attributes)
+
+    if epsg != cloud_epsg:
+        projection.point_cloud_conversion_dataframe(cloud, cloud_epsg, epsg)
+
+    # filter cloud
+    if "mask" in cloud:
+        cloud = cloud[cloud["mask"] == 0]
+
+    if cloud.dropna(subset=["x", "y", "z"]).empty:
+        return None
+
+    # Compute start and size
+    if terrain_region is None:
+        # compute region from cloud
+        xmin = np.nanmin(cloud["x"])
+        xmax = np.nanmax(cloud["x"])
+        ymin = np.nanmin(cloud["y"])
+        ymax = np.nanmax(cloud["y"])
+        # Add margin to be sure every point is rasterized
+        terrain_region = [
+            xmin - radius * resolution,
+            ymin - radius * resolution,
+            xmax + radius * resolution,
+            ymax + radius * resolution,
+        ]
+
+        if terrain_full_roi is not None:
+            # Modify start (used in tiling.roi_to_start_and_size)  [0, 3]
+            # to share the same global grid
+            terrain_region[0] = (
+                terrain_full_roi[0]
+                + np.round(
+                    (terrain_region[0] - terrain_full_roi[0]) / resolution
+                )
+                * resolution
+            )
+            terrain_region[3] = (
+                terrain_full_roi[3]
+                + np.round(
+                    (terrain_region[3] - terrain_full_roi[3]) / resolution
+                )
+                * resolution
+            )
+            # Crop
+            terrain_region = [
+                max(terrain_full_roi[0], terrain_region[0]),
+                max(terrain_full_roi[1], terrain_region[1]),
+                min(terrain_full_roi[2], terrain_region[2]),
+                min(terrain_full_roi[3], terrain_region[3]),
+            ]
+
+            if (
+                terrain_region[0] > terrain_region[2]
+                or terrain_region[1] > terrain_region[3]
+            ):
+                return None
+
+    xstart, ystart, xsize, ysize = tiling.roi_to_start_and_size(
+        terrain_region, resolution
+    )
+
+    if xsize == 0 or ysize == 0:
+        logging.warning("Tile is empty")
+        return None
+
+    if window is None:
+        transform = rio.Affine(*profile["transform"][0:6])
+        row_pix_pos, col_pix_pos = rio.transform.AffineTransformer(
+            transform
+        ).rowcol(xstart, ystart)
+        window = [
+            row_pix_pos,
+            row_pix_pos + ysize,
+            col_pix_pos,
+            col_pix_pos + xsize,
+        ]
+
+        window = cars_dataset.window_array_to_dict(window)
+
+    # Call simple_rasterization
+    raster = rasterization_algo.simple_rasterization_dataset_wrapper(
+        cloud,
+        resolution,
+        epsg,
+        xstart=xstart,
+        ystart=ystart,
+        xsize=xsize,
+        ysize=ysize,
+        sigma=sigma,
+        radius=radius,
+        dsm_no_data=dsm_no_data,
+        texture_no_data=texture_no_data,
+        msk_no_data=msk_no_data,
+        list_computed_layers=list_computed_layers,
+        source_pc_names=source_pc_names,
+        performance_map_classes=performance_map_classes,
+        cloud_global_id=attributes["cloud_id"],
+    )
+
+    # Fill raster
+    attributes = {
+        "color_type": color_dtype,
+        cst.CROPPED_DISPARITY_RANGE: (ocht.get_disparity_range_cropped(cloud)),
+    }
+    if raster is not None:
+        cars_dataset.fill_dataset(
+            raster,
+            saving_info=saving_info,
+            window=window,
+            profile=profile,
+            attributes=attributes,
+            overlaps=None,
+        )
+
+    return raster
+
+
+def raster_final_function(orchestrator, future_object):
+    """
+    Apply function to current object, reading already rasterized data
+
+    :param orchestrator: orchestrator
+    :param future_object: Dataset
+
+    :return: update object
+    """
+    # Get data weights
+    old_weights, _ = orchestrator.get_data(
+        cst.RASTER_WEIGHTS_SUM, future_object
+    )
+    weights = future_object[cst.RASTER_WEIGHTS_SUM].values
+
+    future_object[cst.RASTER_WEIGHTS_SUM].values = np.reshape(
+        rasterization_wrappers.update_weights(old_weights, weights),
+        weights.shape,
+    )
+
+    # Get color type
+    color_type = future_object.attrs["attributes"]["color_type"]
+
+    # Get data dsm
+    for tag in future_object.keys():
+
+        if tag != cst.RASTER_WEIGHTS_SUM:
+
+            if tag in [cst.RASTER_NB_PTS, cst.RASTER_NB_PTS_IN_CELL]:
+                method = "sum"
+            elif tag in [
+                cst.RASTER_FILLING,
+                cst.RASTER_CLASSIF,
+                cst.RASTER_SOURCE_PC,
+            ]:
+                method = "bool"
+            else:
+                method = "basic"
+
+            old_data, nodata_raster = orchestrator.get_data(tag, future_object)
+            current_data = future_object[tag].values
+            if tag == cst.RASTER_COLOR_IMG and np.issubdtype(
+                color_type, np.integer
+            ):
+                current_data = np.round(current_data).astype(color_type)
+            future_object[tag].values = np.reshape(
+                rasterization_wrappers.update_data(
+                    old_data,
+                    current_data,
+                    weights,
+                    old_weights,
+                    nodata_raster,
+                    method=method,
+                ),
+                current_data.shape,
+            )
+
+    return future_object