# -*- python -*-
#
# Copyright INRIA - CIRAD - INRA
#
# Distributed under the Cecill-C License.
# See accompanying file LICENSE.txt or copy at
# http://www.cecill.info/licences/Licence_CeCILL-C_V1-en.html
#
# ==============================================================================
import collections
import json
import os
from pathlib import Path
import numpy
from PIL import Image
from openalea.phenomenal.calibration import (
Chessboard,
Chessboards,
Calibration,
CalibrationSetup,
OldCalibrationCamera,
)
from openalea.phenomenal.mesh import read_ply_to_vertices_faces
from openalea.phenomenal.object import VoxelGrid
# ==============================================================================
def _path_images(data_dir, dtype="bin"):
"""According to the plant number return a dict[id_camera][angle] containing
filename of file.
Parameters
----------
dtype : "bin" or "raw" or "chessboard"
Returns
-------
d : dict of dict of string
dict[id_camera][angle] = filename
"""
data_directory = Path(data_dir) / f"{dtype}"
d = collections.defaultdict(dict)
for camera_dir in data_directory.iterdir():
id_camera = str(camera_dir.stem)
if id_camera in ('side', 'top'):
for filename in (data_directory / id_camera).iterdir():
angle = int(filename.stem)
d[id_camera][angle] = filename
return d
[docs]
def path_bin_images(data_dir):
"""According to the plant number return a dict[id_camera][angle] containing
filename of the binary image.
Returns
-------
d : dict of dict of string
dict[id_camera][angle] = filename
"""
return _path_images(data_dir, dtype="bin")
[docs]
def path_raw_images(data_dir):
"""
According to the plant number return a dict[id_camera][angle] containing
filename of the raw image.
:return: dict[id_camera][angle] of filename
"""
return _path_images(data_dir, dtype="raw")
[docs]
def path_chessboard_images(data_dir):
"""
According to the plant number return a dict[id_camera][angle] containing
filename of the raw image.
:return: dict[id_camera][angle] of filename
"""
return _path_images(data_dir, dtype="chessboard")
[docs]
def raw_images(data_dir):
"""
According to the plant number return a dict[id_camera][angle] of
numpy array of the loader raw image.
:return: dict[id_camera][angle] of loaded RGB image
"""
d = path_raw_images(data_dir)
for id_camera in d:
for angle in d[id_camera]:
d[id_camera][angle] = numpy.asarray(
Image.open(d[id_camera][angle]).convert("RGB"), dtype=numpy.uint8
)
return d
[docs]
def bin_images(data_dir):
"""
According to the plant number return a dict[id_camera][angle] of
numpy array of the loader binary image.
A binary image is a numpy array of uint8 type.
:return: dict[id_camera][angle] of loaded grayscale image
"""
d = path_bin_images(data_dir)
for id_camera in d:
for angle in d[id_camera]:
d[id_camera][angle] = numpy.asarray(
Image.open(d[id_camera][angle]).convert("L"), dtype=numpy.uint8
)
return d
[docs]
def chessboard_images(data_dir):
"""
According to the plant number return a dict[id_camera][angle] of
numpy array of the loader binary image.
A binary image is a numpy array of uint8 type.
:return: dict[id_camera][angle] of loaded grayscale image
"""
d = path_chessboard_images(data_dir)
for id_camera in d:
for angle in d[id_camera]:
d[id_camera][angle] = numpy.asarray(
Image.open(d[id_camera][angle]).convert("RGB")
)
return (d,)
# ==============================================================================
[docs]
def chessboards(data_dir):
"""
According to name_dir return a dict[id_camera] of camera
calibration object
:return: dict[id_camera] of camera calibration object
"""
data_directory = os.path.join(
data_dir, "chessboard/points/"
)
chessboards = []
for id_chessboard in [1, 2]:
chessboards.append(
Chessboard.load(
os.path.join(str(data_directory), f"chessboard_{id_chessboard}.json")
)
)
return chessboards
[docs]
def image_points(data_dir):
"""
According to name_dir return a dict[id_camera] of camera
calibration object
:return: dict[id_camera] of camera calibration object
"""
data_directory = os.path.join(
data_dir, "chessboard/points/"
)
chessboards = {}
keep = [42] + list(range(0, 360, 30))
for id_chessboard in ["target_1", "target_2"]:
chessboard = Chessboard.load(
os.path.join(str(data_directory), f"image_points_{id_chessboard}.json")
)
for rotation in list(chessboard.image_points["side"]):
if rotation not in keep:
chessboard.image_points["side"].pop(rotation)
chessboards[id_chessboard] = chessboard
return chessboards
[docs]
def do_calibration(data_dir):
"""Regenerate calibration of cameras"""
data_directory = os.path.join(data_dir, "calibration")
cbs = dict(zip(("target_1", "target_2"), chessboards(data_dir)))
# add missing info
cb = cbs["target_1"]
cb.facing_angles = {"side": 48, "top": 48}
cb.image_sizes = {"side": (2056, 2454), "top": (2454, 2056)}
cb.check_order()
#
cb = cbs["target_2"]
cb.facing_angles = {"side": 228, "top": 228}
cb.image_sizes = {"side": (2056, 2454), "top": (2454, 2056)}
cb.check_order()
chess_targets = Chessboards(cbs)
image_sizes = chess_targets.image_sizes()
image_resolutions = chess_targets.image_resolutions()
facings = chess_targets.facings()
target_points = chess_targets.target_points()
image_points = chess_targets.image_points()
# distance and inclination of objects
cameras_guess = {"side": (5500, 90), "top": (2500, 0)}
targets_guess = {"target_1": (100, 45), "target_2": (100, 45)}
setup = CalibrationSetup(
cameras_guess,
targets_guess,
image_resolutions,
image_sizes,
facings,
clockwise_rotation=True,
)
cameras, targets = setup.setup_calibration(
reference_camera="side", reference_target="target_1"
)
calibration = Calibration(
targets=targets,
cameras=cameras,
target_points=target_points,
image_points=image_points,
reference_camera="side",
clockwise_rotation=True,
)
calibration.calibrate()
calibration.dump(os.path.join(data_directory, "calibration_cameras.json"))
[docs]
def calibrations(data_dir):
"""
According to name_dir return a dict[id_camera] of camera
calibration object
:return: dict[id_camera] of camera calibration object
"""
data_directory = os.path.join(data_dir, "calibration/")
calibration = {}
for id_camera in ["side", "top"]:
calibration[id_camera] = OldCalibrationCamera.load(
os.path.join(data_directory, f"calibration_camera_{id_camera}.json")
)
return calibration
[docs]
def new_calibrations(data_dir):
"""
According to name_dir return a camera
calibration object
"""
file_name = os.path.join(data_dir, "calibration", "calibration_cameras.json")
return Calibration.load(file_name)
[docs]
def voxel_grid(data_dir, voxels_size=4):
"""
According to the plant number and the voxel size desired return the
voxel_grid of the plant.
:param plant_number: number of the plant desired (int)
:param voxels_size: diameter of each voxel in mm (int)
:return: voxel_grid object
"""
vg = VoxelGrid.read(
str(data_dir / "voxels" / f"{voxels_size}.npz")
)
return vg
# ==============================================================================
[docs]
def tutorial_data_binarization_mask(data_dir):
"""
Return the list of required images to process the notebook tutorial on
binarization. The images are already load with opencv in unchanged format.
images = ["mask_hsv.png", "mask_clean_noise.png", "mask_mean_shift.png"]
:return: list of image
"""
data_directory = os.path.join(data_dir, "mask/")
masks = []
for filename in ["mask_hsv.png", "mask_mean_shift.png"]:
masks.append(numpy.asarray(Image.open(os.path.join(data_directory, filename))))
return masks
# ==============================================================================
[docs]
def synthetic_plant(data_dir, registration_point=(0, 0, 750)):
"""According to name_dir return the mesh plant and skeleton of the
synthetic plant.
Parameters
----------
name_dir : str
Name of the synthetic plant directory
registration_point: 3-tuple, optional
Position of the pot in the scene
Returns
-------
out : vertices, faces, meta_data
"""
filename = os.path.join(data_dir, "synthetic_plant.ply")
vertices, faces, _ = read_ply_to_vertices_faces(filename)
vertices = numpy.array(vertices) * 10 - numpy.array([registration_point])
with open(filename.replace("ply", "json"), "r", encoding="UTF8") as infile:
meta_data = json.load(infile)
return vertices, faces, meta_data
# ==============================================================================
[docs]
def mesh_mccormik_plant(data_dir):
"""According to name_dir return the mesh of plant from the McCormik paper"""
filename = os.path.join(data_dir, "segmentedMesh.ply")
vertices, faces, colors = read_ply_to_vertices_faces(filename)
return vertices, faces, colors
