cars 1.0.0rc1__cp312-cp312-manylinux_2_17_i686.manylinux2014_i686.whl

cars/applications/triangulation/pc_transform.py

@@ -0,0 +1,552 @@
+#!/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 is responsible for the transition between triangulation and
+rasterization steps
+"""
+# pylint: disable=C0302
+
+# Standard imports
+import logging
+from typing import List, Tuple, Union
+
+# Third party imports
+import numpy as np
+import pandas as pd
+import xarray as xr
+
+from cars.applications.dense_matching import dense_matching_wrappers
+
+# CARS imports
+from cars.core import constants as cst
+from cars.core import projection
+
+
+def filter_cloud_with_mask(crop_cloud, crop_terrain_tile_data_msk):
+    """
+    Delete masked points with terrain tile mask
+
+    :param crop_cloud: the point cloud
+    :param crop_terrain_tile_data_msk: terrain tile mask
+    """
+    crop_terrain_tile_data_msk = np.ravel(crop_terrain_tile_data_msk)
+
+    crop_terrain_tile_data_msk_pos = np.nonzero(~crop_terrain_tile_data_msk)
+
+    crop_cloud = np.delete(crop_cloud, crop_terrain_tile_data_msk_pos[0], 0)
+
+    return crop_cloud
+
+
+# pylint: disable=too-many-positional-arguments
+def compute_terrain_msk(
+    dsm_epsg,
+    xmin,
+    xmax,
+    ymin,
+    ymax,
+    margin,
+    epsg,
+    point_cloud,
+    full_x,
+    full_y,
+):
+    """
+    Compute terrain tile msk bounds
+
+    If the point clouds are not in the same referential as the roi,
+    it is converted using the dsm_epsg
+
+    :param dsm_epsg: epsg code for the CRS of the final output raster
+    :param xmin: xmin of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param xmax: xmax of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param ymin: ymin of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param ymax: ymax of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param margin: Margin added for each tile, in meter or degree.
+        (default value: 0)
+    :param epsg: epsg code of the input cloud
+    :param point_cloud: the point cloud
+    :param full_x: point_cloud[X]
+    :param full_y: point_cloud[Y]
+    """
+    if epsg != dsm_epsg:
+        (
+            full_x,
+            full_y,
+        ) = projection.get_converted_xy_np_arrays_from_dataset(
+            point_cloud, dsm_epsg
+        )
+    msk_xmin = np.where(full_x > xmin - margin, True, False)
+    msk_xmax = np.where(full_x < xmax + margin, True, False)
+    msk_ymin = np.where(full_y > ymin - margin, True, False)
+    msk_ymax = np.where(full_y < ymax + margin, True, False)
+    terrain_tile_data_msk = np.logical_and(
+        msk_xmin,
+        np.logical_and(msk_xmax, np.logical_and(msk_ymin, msk_ymax)),
+    )
+    terrain_tile_data_msk_pos = terrain_tile_data_msk.astype(np.int8).nonzero()
+
+    return terrain_tile_data_msk, terrain_tile_data_msk_pos
+
+
+def create_point_cloud_index(cloud_sample):
+    """
+    Create point cloud index from cloud list keys and color inputs
+    """
+    cloud_indexes_with_types = {
+        cst.X: "float64",
+        cst.Y: "float64",
+        cst.Z: "float64",
+    }
+
+    # Add Z_inf Z_sup, and performance maps if computed
+    for key in list(cloud_sample.keys()):
+        if (
+            cst.POINT_CLOUD_LAYER_INF in key
+            or cst.POINT_CLOUD_LAYER_SUP in key
+            or cst.POINT_CLOUD_PERFORMANCE_MAP_ROOT in key
+        ):
+            cloud_indexes_with_types[key] = "float32"
+
+    # Add mask index
+    if cst.EPI_MSK in cloud_sample:
+        cloud_indexes_with_types[cst.POINT_CLOUD_MSK] = "uint8"
+
+    # Add color indexes
+    if cst.EPI_TEXTURE in cloud_sample:
+        band_color = list(cloud_sample.coords[cst.BAND_IM].to_numpy())
+        color_type = "float32"
+        if "color_type" in cloud_sample.attrs:
+            color_type = cloud_sample.attrs["color_type"]
+        for band in band_color:
+            band_index = "{}_{}".format(cst.POINT_CLOUD_CLR_KEY_ROOT, band)
+            cloud_indexes_with_types[band_index] = color_type
+
+    # Add classif indexes
+    if cst.EPI_CLASSIFICATION in cloud_sample:
+        band_classif = list(cloud_sample.coords[cst.BAND_CLASSIF].to_numpy())
+        for band in band_classif:
+            band_index = "{}_{}".format(cst.POINT_CLOUD_CLASSIF_KEY_ROOT, band)
+            cloud_indexes_with_types[band_index] = "boolean"
+
+    # Add filling information indexes
+    if cst.EPI_FILLING in cloud_sample:
+        band_filling = list(cloud_sample.coords[cst.BAND_FILLING].to_numpy())
+        for band in band_filling:
+            band_index = "{}_{}".format(cst.POINT_CLOUD_FILLING_KEY_ROOT, band)
+            cloud_indexes_with_types[band_index] = "uint8"
+
+    # Add ambiguity information index
+    if cst.EPI_AMBIGUITY in cloud_sample:
+        cloud_indexes_with_types[cst.EPI_AMBIGUITY] = "float32"
+
+    return cloud_indexes_with_types
+
+
+# pylint: disable=too-many-positional-arguments
+def add_information_to_cloud(
+    input_cloud, cloud_indexes, bbox, target_cloud, input_array, output_column
+):
+    """
+    Add color information for a current cloud_list item
+
+    :param cloud: source point cloud dataset
+    :type cloud: xr.Dataset
+    :param cloud_indexes: list of band data to extract
+    :type cloud_indexes: list[str]
+    :param bbox: bbox of interest
+    :type bbox: list[int]
+    :param crop_cloud: target flatten point cloud
+    :type crop_cloud: np.array[columns, points]
+    :param input_array: index of input to extract from cloud
+    :type input_array: str
+    :param output_column: index of crop_cloud to fill
+    :type input_array: str
+    """
+    if input_array in input_cloud:
+        full_array = input_cloud[input_array]
+        if len(full_array.shape) == 3:
+            # Array with multiple bands
+            array = full_array[:, bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1]
+            for column_name in cloud_indexes:
+                if output_column in column_name:
+                    band_name = column_name.replace(output_column + "_", "")
+                    band = array.loc[band_name]
+                    index = cloud_indexes.index(column_name)
+                    target_cloud[index, :] = np.ravel(band.values)
+        elif len(full_array.shape) == 2:
+            # Array with single band
+            array = full_array[bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1]
+            index = cloud_indexes.index(output_column)
+            target_cloud[index, :] = np.ravel(array.values)
+
+
+def get_color_type(clouds):
+    """
+    Get color type of the tiles and if the same type.
+
+    :param cloud_list: list of clouds
+    :type cloud_list: xarray Dataset
+
+    :return: color type of the tiles list
+    :rtype: str
+
+    """
+    color_types = []
+    for cloud_id, cloud_item in enumerate(clouds):
+        if cst.EPI_TEXTURE in clouds[cloud_id]:
+            if "color_type" in cloud_item.attrs:
+                color_types.append(cloud_item.attrs["color_type"])
+    if color_types:
+        color_type_set = set(color_types)
+        if len(color_type_set) > 1:
+            logging.warning("The tiles colors don't have the same type.")
+        return color_types[0]
+
+    return None
+
+
+def filter_cloud(
+    cloud: pd.DataFrame,
+    index_elt_to_remove: List[int],
+    filtered_elt_pos: bool = False,
+) -> Tuple[pd.DataFrame, Union[None, pd.DataFrame]]:
+    """
+    Filter all points of the cloud DataFrame
+    which index is in the index_elt_to_remove list.
+
+    If filtered_elt_pos is set to True, the information of the removed elements
+    positions in their original epipolar images are returned.
+
+    To do so the cloud DataFrame has to be build
+    with the 'with_coords' option activated.
+
+    :param cloud: combined cloud
+        as returned by the create_combined_cloud function
+    :param index_elt_to_remove: indexes of lines
+        to filter in the cloud DataFrame
+    :param filtered_elt_pos: if filtered_elt_pos is set to True,
+        the removed points positions in their original epipolar images are
+        returned, otherwise it is set to None
+    :return: Tuple composed of the filtered cloud DataFrame and
+        the filtered elements epipolar position information
+        (or None for the latter if filtered_elt_pos is set to False
+        or if the cloud Dataframe has not been build with with_coords option)
+    """
+    if filtered_elt_pos and not (
+        cst.POINT_CLOUD_COORD_EPI_GEOM_I in cloud.columns
+        and cst.POINT_CLOUD_COORD_EPI_GEOM_J in cloud.columns
+        and cst.POINT_CLOUD_ID_IM_EPI in cloud.columns
+    ):
+        logging.warning(
+            "In filter_cloud: the filtered_elt_pos has been activated but "
+            "the cloud Datafram has not been build with option with_coords. "
+            "The positions cannot be retrieved."
+        )
+        filtered_elt_pos = False
+
+    # retrieve removed points position in their original epipolar images
+    if filtered_elt_pos:
+        labels = [
+            cst.POINT_CLOUD_COORD_EPI_GEOM_I,
+            cst.POINT_CLOUD_COORD_EPI_GEOM_J,
+            cst.POINT_CLOUD_ID_IM_EPI,
+        ]
+
+        removed_elt_pos_infos = cloud.loc[
+            cloud.index.values[index_elt_to_remove], labels
+        ].values
+
+        removed_elt_pos_infos = pd.DataFrame(
+            removed_elt_pos_infos, columns=labels
+        )
+    else:
+        removed_elt_pos_infos = None
+
+    # remove points from the cloud
+    cloud = cloud.drop(index=cloud.index.values[index_elt_to_remove])
+
+    return cloud, removed_elt_pos_infos
+
+
+# pylint: disable=too-many-positional-arguments
+def depth_map_dataset_to_dataframe(  # noqa: C901
+    cloud_dataset: xr.Dataset,
+    dsm_epsg: int,
+    xmin: float = None,
+    xmax: float = None,
+    ymin: int = None,
+    ymax: int = None,
+    margin: float = 0,
+    with_coords: bool = False,
+) -> Tuple[pd.DataFrame, int]:
+    """
+    Combine a list of clouds (and their colors) into a pandas dataframe
+    structured with the following labels:
+
+        - if no colors in input and no mask data present in cloud_list datasets:
+            labels=[cst.X, cst.Y, cst.Z] \
+            The combined cloud has x, y, z columns
+
+        - if no colors in input and mask data present in cloud_list datasets:
+            labels=[cst.X, cst.Y, cst.Z, cst.POINT_CLOUD_MSK]\
+            The mask values are added to the dataframe.
+
+        - if colors are set in input and mask data are present \
+            in the cloud_list datasets:
+           labels=[cst.X, cst.Y, cst.Z, cst.POINT_CLOUD_MSK,\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"0",\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"1",\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"2"]\
+            Color channels information are added to the dataframe.
+
+        - if colors in input, mask data present in the cloud_list datasets and\
+            the with_coords option is activated:
+             labels=[cst.X, cst.Y, cst.Z, cst.POINT_CLOUD_MSK,\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"0",\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"1",\
+                     cst.POINT_CLOUD_CLR_KEY_ROOT+"2"\
+                     cst.POINT_CLOUD_COORD_EPI_GEOM_I,\
+                     cst.POINT_CLOUD_COORD_EPI_GEOM_J,\
+                     cst.POINT_CLOUD_ID_IM_EPI]\
+            The pixel position of the xyz point in the original epipolar\
+            image (coord_epi_geom_i, coord_epi_geom_j) are added\
+            to the dataframe along with the index of its original cloud\
+            in the cloud_list input.
+        - if confidence intervals on Z in input, then\
+            [cst.Z_INF, cst.Z_SUP] are also added to the labels
+
+
+    :param dsm_epsg: epsg code for the CRS of the final output raster
+    :param xmin: xmin of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param xmax: xmax of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param ymin: ymin of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param ymax: ymax of the rasterization grid
+        (if None, the whole clouds are combined)
+    :param margin: Margin added for each tile, in meter or degree.
+        (default value: 0)
+    :param with_coords: Option enabling the adding to the combined cloud
+        of information of each point to retrieve their positions
+        in the original epipolar images
+    :return: Tuple formed with the combined clouds and color
+        in a single pandas dataframe and the epsg code
+    """
+    epsg = int(cloud_dataset.attrs[cst.EPSG])
+
+    # Compute margin/roi and final number of data to add to the combined cloud
+    roi = (
+        xmin is not None
+        and xmax is not None
+        and ymin is not None
+        and ymax is not None
+    )
+
+    # Create point cloud index
+    cloud_indexes_with_types = create_point_cloud_index(cloud_dataset)
+
+    # Add coords
+    if with_coords:
+        cloud_indexes_with_types.update(
+            {
+                cst.POINT_CLOUD_COORD_EPI_GEOM_I: "uint16",
+                cst.POINT_CLOUD_COORD_EPI_GEOM_J: "uint16",
+            }
+        )
+
+    cloud_indexes = list(cloud_indexes_with_types.keys())
+
+    # crop point cloud if is not created from tif depth maps
+    if (
+        cst.EPI_MARGINS in cloud_dataset.attrs
+        and cst.ROI in cloud_dataset.attrs
+    ):
+        ref_roi, _, _ = dense_matching_wrappers.compute_cropped_roi(
+            cloud_dataset.attrs[cst.EPI_MARGINS],
+            0,
+            cloud_dataset.attrs[cst.ROI],
+            cloud_dataset.sizes[cst.ROW],
+            cloud_dataset.sizes[cst.COL],
+        )
+        cloud_dataset = cloud_dataset.isel(
+            row=slice(ref_roi[1], ref_roi[3]),
+            col=slice(ref_roi[0], ref_roi[2]),
+        )
+
+    full_x = cloud_dataset[cst.X].values
+    full_y = cloud_dataset[cst.Y].values
+    full_z = cloud_dataset[cst.Z].values
+
+    # get mask of points inside the roi (plus margins)
+    if roi:
+        # Compute terrain tile bounds
+        # if the point clouds are not in the same referential as the roi,
+        # it is converted using the dsm_epsg
+        (
+            terrain_tile_data_msk,
+            terrain_tile_data_msk_pos,
+        ) = compute_terrain_msk(
+            dsm_epsg,
+            xmin,
+            xmax,
+            ymin,
+            ymax,
+            margin,
+            epsg,
+            cloud_dataset,
+            full_x,
+            full_y,
+        )
+
+        # if the point clouds are not in the same referential as the roi,
+        # retrieve the initial values
+        if epsg != dsm_epsg:
+            full_x = cloud_dataset[cst.X].values
+            full_y = cloud_dataset[cst.Y].values
+
+        # if no point is found, return Empty cloud
+        if terrain_tile_data_msk_pos[0].shape[0] == 0:
+            logging.error("No points found to transform")
+            return pd.DataFrame(columns=cloud_indexes), epsg
+
+        # get useful data bounding box
+        bbox = [
+            np.min(terrain_tile_data_msk_pos[0]),
+            np.min(terrain_tile_data_msk_pos[1]),
+            np.max(terrain_tile_data_msk_pos[0]),
+            np.max(terrain_tile_data_msk_pos[1]),
+        ]
+    else:
+        bbox = [0, 0, full_y.shape[0] - 1, full_y.shape[1] - 1]
+
+    # add (x, y, z) information to the current cloud
+    crop_x = full_x[bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1]
+    crop_y = full_y[bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1]
+    crop_z = full_z[bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1]
+
+    flatten_cloud = np.zeros(
+        (
+            len(cloud_indexes),
+            (bbox[2] - bbox[0] + 1) * (bbox[3] - bbox[1] + 1),
+        )
+    )
+    flatten_cloud[cloud_indexes.index(cst.X), :] = np.ravel(crop_x)
+    flatten_cloud[cloud_indexes.index(cst.Y), :] = np.ravel(crop_y)
+    flatten_cloud[cloud_indexes.index(cst.Z), :] = np.ravel(crop_z)
+
+    # add additional information to point cloud
+    arrays_to_add_to_point_cloud = [
+        (cst.EPI_TEXTURE, cst.POINT_CLOUD_CLR_KEY_ROOT),
+        (cst.EPI_MSK, cst.POINT_CLOUD_MSK),
+        (cst.EPI_CLASSIFICATION, cst.POINT_CLOUD_CLASSIF_KEY_ROOT),
+        (cst.EPI_FILLING, cst.POINT_CLOUD_FILLING_KEY_ROOT),
+    ]
+
+    # Add layer inf and sup
+    for array_name in cloud_dataset:
+        if cst.POINT_CLOUD_LAYER_SUP_OR_INF_ROOT in array_name:
+            arrays_to_add_to_point_cloud.append((array_name, array_name))
+
+    # add performance map
+    for array_name in cloud_dataset:
+        if cst.POINT_CLOUD_PERFORMANCE_MAP_ROOT in array_name:
+            arrays_to_add_to_point_cloud.append((array_name, array_name))
+
+    # add ambiguity layer, drop confidence_* layers
+    for array_name in cloud_dataset:
+        if (
+            cst.EPI_AMBIGUITY in array_name
+            and cst.EPI_CONFIDENCE_KEY_ROOT not in array_name
+        ):
+            arrays_to_add_to_point_cloud.append((array_name, array_name))
+
+    # add denoising info layers
+    for array_name in cloud_dataset:
+        if cst.EPI_DENOISING_INFO_KEY_ROOT in array_name:
+            arrays_to_add_to_point_cloud.append((array_name, array_name))
+
+    for input_band, output_column in arrays_to_add_to_point_cloud:
+        add_information_to_cloud(
+            cloud_dataset,
+            cloud_indexes,
+            bbox,
+            flatten_cloud,
+            input_band,
+            output_column,
+        )
+
+    # add the original image coordinates information to the current cloud
+    if with_coords:
+        coords_line = np.linspace(bbox[0], bbox[2], bbox[2] - bbox[0] + 1)
+        coords_col = np.linspace(bbox[1], bbox[3], bbox[3] - bbox[1] + 1)
+        coords_col, coords_line = np.meshgrid(coords_col, coords_line)
+
+        flatten_cloud[
+            cloud_indexes.index(cst.POINT_CLOUD_COORD_EPI_GEOM_I), :
+        ] = np.ravel(coords_line)
+        flatten_cloud[
+            cloud_indexes.index(cst.POINT_CLOUD_COORD_EPI_GEOM_J), :
+        ] = np.ravel(coords_col)
+
+    # Transpose point cloud
+    flatten_cloud = flatten_cloud.transpose()
+
+    # remove masked data (pandora + out of the terrain tile points)
+    crop_terrain_tile_data_msk = (
+        cloud_dataset[cst.POINT_CLOUD_CORR_MSK].values[
+            bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1
+        ]
+        == 255
+    )
+
+    if roi:
+        crop_terrain_tile_data_msk = np.logical_and(
+            crop_terrain_tile_data_msk,
+            terrain_tile_data_msk[bbox[0] : bbox[2] + 1, bbox[1] : bbox[3] + 1],
+        )
+
+    flatten_cloud = filter_cloud_with_mask(
+        flatten_cloud, crop_terrain_tile_data_msk
+    )
+
+    # Remove points with nan values on X, Y or Z
+    xyz_indexes = np.array(
+        [
+            cloud_indexes.index(cst.X),
+            cloud_indexes.index(cst.Y),
+            cloud_indexes.index(cst.Z),
+        ]
+    )
+    flatten_cloud = flatten_cloud[
+        ~np.any(np.isnan(flatten_cloud[:, xyz_indexes]), axis=1)
+    ]
+
+    pd_cloud = pd.DataFrame(flatten_cloud, columns=cloud_indexes)
+    pd_cloud = pd_cloud.astype(cloud_indexes_with_types)
+
+    return pd_cloud, epsg