def test_graph():
plant_number = 1
voxels_size = 16
voxel_grid = phm_data.voxel_grid(plant_number=plant_number,
voxels_size=voxels_size)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
def test_graph_1():
voxels_position = [(0, 0, 0),
(0, 0, 1),
(0, 1, 0),
(1, 1, 1),
(1, 1, 2),
(2, 1, 2),
(1, 2, 1),
(2, 2, 2),
(5, 5, 5)]
voxels_size = 1
voxel_grid = phm_obj.VoxelGrid(voxels_position, voxels_size)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
print("Number of nodes : {nb_nodes}".format(nb_nodes=len(graph.nodes())))
assert len(voxels_position) == len(graph.nodes())
for position in voxels_position:
assert graph.has_node(position)
l = networkx.dijkstra_path_length(graph,
source=(0, 0, 0),
target=(2, 2, 2),
weight="weight")
ll = numpy.linalg.norm(numpy.array((0, 0, 0)) - numpy.array((2, 2, 2)))
assert l == ll
def test_maize():
plant_number = 3
voxels_size = 4
voxel_grid = phm_data.voxel_grid(plant_number=plant_number,
voxels_size=voxels_size)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
vms = phm_seg.maize_segmentation(voxel_skeleton, graph)
vmsi = phm_seg.maize_analysis(vms)
def test_maize():
voxel_grid = phm_data.random_voxel_grid(voxels_size=32)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
vms = phm_seg.maize_segmentation(voxel_skeleton, graph)
vmsi = phm_seg.maize_analysis(vms)
# Write
filename = 'tmp.json'
vms.write_to_json_gz(filename)
vms = phm_obj.VoxelSegmentation.read_from_json_gz(filename)
os.remove(filename)
def test_time_graph():
voxel_grid = phm_data.random_voxel_grid(voxels_size=16)
number_of_loop = 2
best_time = float('inf')
for i in range(number_of_loop):
t0 = time.time()
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
best_time = min(best_time, float(time.time() - t0))
print("{number_of_loop} loop, best of {number_of_loop}: "
"{best_time}s per loop".format(best_time=best_time,
number_of_loop=number_of_loop))
def test_running():
plant_number = 1
voxels_size = 16
bin_images = phm_data.bin_images(plant_number=plant_number)
calibrations = phm_data.calibrations(plant_number=plant_number)
voxel_grid = phm_data.voxel_grid(plant_number=plant_number,
voxels_size=voxels_size)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
image_projection = list()
for angle in [0, 120, 270]:
projection = calibrations["side"].get_projection(angle)
image_projection.append((bin_images["side"][angle], projection))
voxel_skeleton_reduced = phm_seg.segment_reduction(voxel_skeleton,
image_projection,
required_visible=1,
nb_min_pixel=100)
def test_running():
bin_images = phm_data.bin_images(plant_1_dir)
calibrations = phm_data.calibrations(plant_1_dir)
voxel_grid = phm_data.voxel_grid(data_dir, 1, 32)
# Load images binarize
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
image_projection = list()
for angle in [0, 120, 270]:
projection = calibrations["side"].get_projection(angle)
image_projection.append((bin_images["side"][angle], projection))
voxel_skeleton = phm_seg.segment_reduction(
voxel_skeleton, image_projection,
required_visible=1,
nb_min_pixel=100)
def test_time_graph():
plant_number = 1
voxels_size = 16
voxel_grid = phm_data.voxel_grid(plant_number=plant_number,
voxels_size=voxels_size)
# print(voxel_grid.voxels_position)
voxels_size = int(voxel_grid.voxels_size)
voxels_position = map(tuple, list(voxel_grid.voxels_position))
voxel_grid = phm_obj.VoxelGrid(voxels_position, voxels_size)
number_of_loop = 2
best_time = float('inf')
for i in range(number_of_loop):
t0 = time.time()
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
best_time = min(best_time, float(time.time() - t0))
print("{number_of_loop} loop, best of {number_of_loop}: "
"{best_time}s per loop".format(best_time=best_time,
number_of_loop=number_of_loop))
def test_graph():
voxel_grid = phm_data.random_voxel_grid(voxels_size=16)
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
def segmentation(row):
bin_images = dc.get_bin_images(row)
calibrations = dc.get_calibrations(row)
refs_angle_list = [_routine_select_ref_angle(bin_images["side"])]
image_views = list()
for id_camera in bin_images:
for angle in bin_images[id_camera]:
projection = calibrations[id_camera].get_projection(angle)
image_ref = None
if id_camera == "side" and angle in refs_angle_list:
image_ref = bin_images[id_camera][angle]
inclusive = False
if id_camera == "top":
inclusive = True
image_views.append(
phm_obj.ImageView(bin_images[id_camera][angle],
projection,
inclusive=inclusive,
image_ref=image_ref))
# ### 2.2 Do multi-view reconstruction
voxels_size = 16 # mm
error_tolerance = 0
voxel_grid = phm_mvr.reconstruction_3d(image_views,
voxels_size=voxels_size,
error_tolerance=error_tolerance)
#phm_display_notebook.show_voxel_grid(voxel_grid, size=1)
# ## 4. Skeletonization
#
# ### 4.1 Compute Graph and the skeleton
# Compute the skeleton of the plant voxelgrid
# In[12]:
graph = phm_seg.graph_from_voxel_grid(voxel_grid, connect_all_point=True)
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
# ### 2.3 Clean skeleton (filter small elements)
# In[14]:
# Select images
image_projection = list()
for angle in bin_images["side"]:
projection = calibrations["side"].get_projection(angle)
image_projection.append((bin_images["side"][angle], projection))
voxel_skeleton_denoised = phm_seg.segment_reduction(voxel_skeleton,
image_projection,
required_visible=5,
nb_min_pixel=400)
#phm_display_notebook.show_skeleton(voxel_skeleton_denoised, size=1)
# ## 5. Maize Segmentation
#
# ### 5.1 Process
#
# Voxel skeleton are segmented in several label "stem", "mature_leaf", "growing_leaf" and "unknow"
# In[16]:
vms = phm_seg.maize_segmentation(voxel_skeleton_denoised, graph)
vmsi = phm_seg.maize_analysis(vms)
return vmsi
#!/usr/bin/env python import sys import openalea.phenomenal.display as phm_display import openalea.phenomenal.mesh as phm_mesh import openalea.phenomenal.object as phm_obj import openalea.phenomenal.segmentation as phm_seg from funcs import * voxel_grid = phm_obj.VoxelGrid.read(sys.argv[1]) graph = phm_seg.graph_from_voxel_grid(voxel_grid, connect_all_point=True) voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
import openalea.phenomenal.segmentation as phm_seg
# import openalea.phenomenal.display as phm_display
# ==============================================================================
plant_number = 2
voxels_size = 8
bin_images = phm_data.bin_images(plant_number=plant_number)
calibrations = phm_data.calibrations(plant_number=plant_number)
voxel_grid = phm_data.voxel_grid(plant_number=plant_number,
voxels_size=voxels_size)
init_start = time.time()
# ==============================================================================
start = time.time()
graph = phm_seg.graph_from_voxel_grid(voxel_grid)
print("time processing , graph_from_voxel_grid : {}".format(time.time() -
start))
start = time.time()
voxel_skeleton = phm_seg.skeletonize(voxel_grid, graph)
print("time processing , skeletonize : {}".format(time.time() - start))
start = time.time()
image_projection = list()
for angle in range(0, 360, 30):
projection = calibrations["side"].get_projection(angle)
image_projection.append((bin_images["side"][angle], projection))
print("time processing , image_projection : {}".format(time.time() - start))
print("len(voxel_skeleton_reduced.segments) : {}".format(
