-
Notifications
You must be signed in to change notification settings - Fork 0
/
pcd_single_thread.py
523 lines (427 loc) · 16.9 KB
/
pcd_single_thread.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
import copy
import gc
import math
import os
import re
import time
# import cv2
import pandas as pd
# from skimage import measure
import numpy
import open3d as o3d
import numpy as np
import random
import itertools
import matplotlib.pyplot as plt
# 把点云找到边角建立坐标系归0
def zero(pcd_in):
np_points = np.asarray(pcd_in.points) - np.asarray(pcd_in.points).min(0)
zero_pcd = o3d.geometry.PointCloud()
zero_pcd.points = o3d.utility.Vector3dVector(np_points)
zero_pcd.colors = pcd_in.colors
return zero_pcd
# 读取bin 转为pcd格式
#无用
def read_point_cloud_bin(bin_path):
data = np.fromfile(bin_path, dtype=np.float32)
# format:
N, D = data.shape[0] // 6, 6
point_cloud_with_normal = np.reshape(data, (N, D))
point_cloud = o3d.geometry.PointCloud()
point_cloud.points = o3d.utility.Vector3dVector(point_cloud_with_normal[:, 0:3])
point_cloud.colors = o3d.utility.Vector3dVector(point_cloud_with_normal[:, 3:6])
return point_cloud
# shape: 划分8,0.8,3
# height:提取区域max高度
# g_scale:计算计算绿色分量 G/R+G+B
# h / 2--定义冠层高度
def cal_para(pcd_in, para):
max_xyz = np.asarray(pcd_in.points).max(0)
print("区域大小为 : {}".format(max_xyz))
avg_x = (max_xyz[0] - 0.3) / para[0]
count_y = int(max_xyz[1] // para[1]) # 向下取整
print("max_xyz : {}".format(max_xyz))
print("avg_x : {0},count_y : {1}".format(avg_x, count_y))
shape = (count_y, para[0])
height = np.zeros(shape)
volume = np.zeros(shape)
surface = np.zeros(shape)
inclination = np.zeros(shape)
g_scale = np.zeros(shape)
canopy_ratio = np.zeros(shape)
leaf_roll = np.zeros(shape)
points = np.zeros(shape)
# 划分y--长
for i in range(count_y):
np_points = np.asarray(pcd_in.points)
np_colors = np.asarray(pcd_in.colors)
bool_y = np.logical_and(np_points[:, 1] >= i * 0.8+0.1, np_points[:, 1] < (i + 1) * 0.8-0.1)
y = np_points[bool_y]
colors_y = np_colors[bool_y]
i += 1
# 划分x--宽
for j in range(para[0]):
if j < (para[0] / 2):
bool_xy = np.logical_and(y[:, 0] >= j * avg_x, y[:, 0] < (j + 1) * avg_x)
else:
bool_xy = np.logical_and(y[:, 0] >= j * avg_x + 0.3, y[:, 0] < (j + 1) * avg_x + 0.3)
# print("j * avg_x : {0}".format(j * avg_x))
xy = y[bool_xy]
# print("xy.shape : {0}".format(xy.shape))
colors_xy = colors_y[bool_xy]
# 保存小区
# if i == 6 and j == 7:
# temp_pcd = np_trans2pcd(xy, colors_xy)
# o3d.io.write_point_cloud("temp/57.pcd", temp_pcd)
# # 保存小区
# if i == 6 and j == 6:
# temp_pcd = np_trans2pcd(xy, colors_xy)
# o3d.io.write_point_cloud("temp/56.pcd", temp_pcd)
pcd_xy = np_trans2pcd(xy, colors_xy)
print("i j : {0},{1} ".format(i, j))
h = cal_height(pcd_xy, avg_x, 0.8, 5, h_ratio=0.1) # 计算最高值0.1以上的高度
height[i - 1, j] = h
bool_xyz = xy[:, 2] > 0.6 * h # h/2以上认为是冠层--超参
np_colors_z = colors_xy[bool_xyz] # 高度达到h/2以上的点
np_points_z = xy[bool_xyz]
pcd_z = np_trans2pcd(np_points_z, np_colors_z)
# 计算参数
# g_scale[i - 1, j] = np_colors_z[:, 1].sum() / np_colors_z.sum()
pcd_z = uniform_sample(pcd_z, np_points_z.shape[0], 50000) # 降采样到5w到10w之间
# 保存第一行
# if i <=1:
# o3d.io.write_point_cloud("pcd/{0}.pcd".format(j),pcd_z)
# #降采样--点少于一定程度---随机下采样效果差
# if np_points_z.shape[0] >= 200000:
# np_points_sample_z, sample_index = random_sample(np_points_z, 200000)
# np_colors_sample_z = np_colors_z[sample_index]
# pcd_z = np_trans2pcd(np_points_sample_z, np_colors_sample_z)
# #均值下采样
# if np_points_z.shape[0] >= 100000:
# pcd_z = o3d.geometry.PointCloud.uniform_down_sample(pcd_z, 4)
#
#保存小区
# if i == 1 and j == 0:
# o3d.io.write_point_cloud("temp/00z.pcd", pcd_z)
# 计算体素后的体积--参数pcd,vexel_size
# volume_ij = cal_volume(pcd_z, voxel_size_in=0.002)
# volume[i - 1, j] = volume_ij
inclination[i - 1, j] = cal_mesh(pcd_z, voxel_size=0.002, alpha=0.01)
# leaf_roll[i - 1, j], canopy_ratio[i - 1, j] = cal_roll_canopy(pcd_z, size=0.001)
j += 1
return height, inclination
# # 去除空洞
# def remove_small_points(image, threshold_point):
# # img = cv2.imread(image, 0) # 输入的二值图像
# img_label, num = measure.label(image, neighbors=8, return_num=True) # 输出二值图像中所有的连通域
# props = measure.regionprops(img_label) # 输出连通域的属性,包括面积等
#
# resMatrix = np.zeros(img_label.shape)
# for i in range(1, len(props)):
# if props[i].area > threshold_point:
# tmp = (img_label == i + 1).astype(np.uint8)
# resMatrix += tmp # 组合所有符合条件的连通域
# resMatrix *= 255
# return resMatrix
# 计算卷叶程度:面积/周长
# 计算冠层覆盖率
# def cal_roll_canopy(pcd, size=0.001, min_area=200, max_area=5000):
# np_2d = sorted(pcd)
# shape = np.ceil(np_2d.max(0) / size)
# shape = shape.astype(int)
#
# bin_image = np.zeros(tuple(shape), dtype="int32")
# i, j, k = 0, 0, 0
# for x in range(shape[0]):
# j = 0
#
# while x * size <= np_2d[j][0] < (x + 1) * size:
# j += 1
# if j >= np_2d.shape[0]:
# break
# sild = np_2d[0:j]
# sild = sild[sild[:, 1].argsort()]
#
# if j != 0:
# np_2d = np.delete(np_2d, np.arange(j), axis=0)
#
# for y in range(shape[1]):
# k = 0
# if sild.shape[0] == 0:
# break
#
# while y * size <= sild[k][1] < (y + 1) * size:
# bin_image[x, y] = 1
# k += 1
# if k >= sild.shape[0]:
# break
#
# if k != 0:
# sild = np.delete(sild, np.arange(k), axis=0)
#
# bin_image = bin_image * 255
# bin_image.dtype = np.uint8
# kernel = np.asarray([[0, 1, 0],
# [1, 1, 1],
# [0, 1, 0]])
# kernel1 = np.ones((1, 1), dtype="int32")
# kernel2 = np.ones((2, 2), dtype="int32")
# kernel3 = np.ones((3, 3), dtype="int32")
# kernel4 = np.ones((4, 4), dtype="int32")
# kernel_cov = np.ones((3, 3), np.float32) / 10
#
# img = cv2.filter2D(bin_image, -1, kernel3)
# img = cv2.dilate(img, kernel2, iterations=1)
# img = cv2.erode(img, kernel2, iterations=1)
#
# img = img / 255
# # 冠层覆盖率
# canopy_ratio = img.sum() / img.size
# img = remove_small_points(img, 500)
# # img = cv2.erode(img, kernel2, iterations=1)
# # img = cv2.filter2D(img, -1, kernel3)
# img = np.asarray(img, dtype=np.uint8)
# # 获取轮廓contours
# contours, layer_num = cv2.findContours(img, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
#
# list_area = np.asarray([])
# list_len = np.asarray([])
# for i in range(len(contours)):
# area = cv2.contourArea(contours[i])
# if area <= min_area:
# continue
# elif area >= max_area:
# continue
# length = cv2.arcLength(contours[i], True)
# list_area = np.append(list_area, area)
# list_len = np.append(list_len, length)
# # print("area:{}".format(area))
# # print("卷叶:{}".format(area / length))
# mean_leaf_roll = np.nanmean(list_area / list_len)
#
# return mean_leaf_roll, canopy_ratio
# xy归0后排序
def sorted(pcd):
np_points = np.asarray(pcd.points)
np_xy = np_points[:, 0:2]
np_2d = np_xy - np_xy.min(0)
idex = np.lexsort([np_2d[:, 1], np_2d[:, 0]])
sorted_num_all = np_2d[idex, :]
return sorted_num_all
# 计算冠层覆盖率,可输出图片,size=0.005 pixel大约160*160
def cal_canopy_ratio(pcd, size):
np_2d = sorted(pcd)
shape = np.ceil(np_2d.max(0) / size)
shape = shape.astype(int)
bin_image = np.zeros(tuple(shape), dtype="int32")
i, j, k = 0, 0, 0
for x in range(shape[0]):
j = 0
while x * size <= np_2d[j][0] < (x + 1) * size:
j += 1
if j >= np_2d.shape[0]:
break
sild = np_2d[0:j]
sild = sild[sild[:, 1].argsort()]
if j != 0:
np_2d = np.delete(np_2d, np.arange(j), axis=0)
for y in range(shape[1]):
k = 0
if sild.shape[0] == 0:
break
while y * size <= sild[k][1] < (y + 1) * size:
bin_image[x, y] = 1
k += 1
if k >= sild.shape[0]:
break
if k != 0:
sild = np.delete(sild, np.arange(k), axis=0)
canopy_ratio = bin_image.sum() / bin_image.size
# #保存图片
# bin_image = bin_image*255
# cv2.imwrite("pcd/img.jpg",bin_image)/
return canopy_ratio
# 均值下采样 到5w点左右
def uniform_sample(pcd_in, count, min):
pcd_in_count = int(re.sub(r'\D', "", str(pcd_in)))
for i in range(2, pcd_in_count // min + 1):
if i * min <= count < (i + 1) * min:
pcd_sample = pcd_in.uniform_down_sample(i)
return pcd_sample
return pcd_in
# 随机下采样到 k 个点,--可返回结果 和 index
def random_sample(np_points, k):
np_choice = np.arange(np_points.shape[0])
index = np.random.choice(np_choice, size=5)
choice_points = np_points[index]
return choice_points, index
# 给一个向量计算向量与Z轴夹角,0 1 2--对应x y z坐标
def cal_theta(array):
if array[2] < 0:
z = -array[2]
else:
z = array[2]
return math.acos(z / math.sqrt((array ** 2).sum())) * 180 / math.pi
# return math.asin(math.sqrt(array[0]*array[0] + array[1]*array[1]) / math.sqrt((array ** 2).sum())) * 180 / math.pi
# return math.acos( array[2] / math.sqrt((array ** 2).sum())) * 180 / math.pi
# 计算mesh化后的 表面积和叶倾角 voxel_size 0.005 alpha = 0.008--
# 0.002 0.009
def cal_mesh(pcd_in, voxel_size, alpha):
pcd_down = pcd_in.voxel_down_sample(voxel_size)
mesh = o3d.geometry.TriangleMesh.create_from_point_cloud_alpha_shape(pcd_down, alpha)
# surface_area = mesh.get_surface_area()
o3d.geometry.TriangleMesh.compute_triangle_normals(mesh)
np_mesh_vertices = np.asarray(mesh.vertices)
np_triangles = np.asarray(mesh.triangles)
np_normals = np.asarray(mesh.triangle_normals)
S = np.zeros(np_normals.shape[0], dtype="double")
leaf_inclination = np.zeros(np_normals.shape[0], dtype="double")
theta = np.zeros(np_normals.shape[0])
for i in range(np_triangles.shape[0]):
triangle = np_mesh_vertices[np_triangles[i]]
AB = triangle[0] - triangle[2]
AC = triangle[1] - triangle[2]
s = np.array([AB[1] * AC[2] - AB[2] * AC[1], AB[2] * AC[0] - AB[0] * AC[2], AB[0] * AC[1] - AB[1] * AC[0]])
S[i] = 0.5 * math.sqrt((s ** 2).sum())
theta[i] = cal_theta(np_normals[i])
if theta[i] < 0:
theta[i] = - theta[i]
leaf_inclination[i] = theta[i] * S[i]
dip_angle = leaf_inclination.sum() / S.sum()
# print("叶倾角theta {0}".format(leaf_inclination.sum() / S.sum()))
# return surface_area, dip_angle
return dip_angle
# 计算高度
def cal_height(pcd, avg_x, avg_y, size, h_ratio=0.1):
height = np.zeros((size, size))
np_points = np.asarray(pcd.points)
min = np.min(np_points, 0)
h_max = np.max(np_points[:, 2])
np_points[:, 0] = np_points[:, 0] - min[0]
np_points[:, 1] = np_points[:, 1] - min[1]
len_x = avg_x / size
len_y = avg_y / size
np_points = np_points[np_points[:, 0].argsort()] # 对x排序
i = 0
j = 0
for x in range(size):
i = 0
if np_points.shape[0] == 0:
break
while x * len_x <= np_points[i][0] < (x + 1) * len_x:
i += 1
if i >= np_points.shape[0]:
break
sild = np_points[0:i]
sild = sild[sild[:, 1].argsort()]
# sild中个数不为0
if i != 0:
np_points = np.delete(np_points, np.arange(i), axis=0)
for y in range(size):
j = 0
if sild.shape[0] == 0:
break
while y * len_y <= sild[j][1] < (y + 1) * len_y:
j += 1
if j >= sild.shape[0]:
break
if j != 0:
temp = sild[0:j]
if np.max(temp[:, 2]) >= h_ratio * h_max:
height[x][y] = np.max(temp[:, 2])
sild = np.delete(sild, np.arange(j), axis=0)
else:
height[x][y] = 0
if np.any(height == 0):
cnt_array = np.where(height, 0, 1)
count = cnt_array.sum()
mean = height.sum() / (size ** 2 - count)
else:
mean = height.mean()
return mean
# 把numpy转pcd
def np_trans2pcd(np_points, np_colors):
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(np_points)
pcd.colors = o3d.utility.Vector3dVector(np_colors)
return pcd
# 计算体素化后的体积
def cal_volume(pcd, voxel_size_in):
voxel_grid = o3d.geometry.VoxelGrid.create_from_point_cloud(pcd,
voxel_size=voxel_size_in)
voxel_grid_count = int(re.sub(r'\D', "", str(voxel_grid)))
volume = voxel_grid_count * voxel_size_in ** 2
return volume
# 计算每一行 所有x区域中的平均高度
def avg_height_x(height):
return height.sum(1) / height.shape[1]
# 计算每一列y中的平均高度
def avg_height_y(height):
return height.sum(0) / height.shape[0]
# 计算总体平均高度
def avg_height(height):
return height.mean()
# 计算绿色分量 G/R+G+B
def cal_g_scale(pcd_in, height):
np_points = np.asarray(pcd_in.points)
# 矩阵左右互换,和实际位置对应
def reverse_matrix(matrix):
return numpy.flip(matrix, 1)
# 可视化矩阵
def plot_matrix(height, name):
# 这里是创建一个数据
height = np.around(height, 3)
y = ["{}".format(i) for i in range(0, height.shape[0])]
x = ["{}".format(i) for i in range(0, height.shape[1])]
# 这里是创建一个画布
fig, ax = plt.subplots(figsize=(7, 14), dpi=400) # 5/27
im = ax.imshow(height, cmap="YlGn") # "YlGn"--绿色,Reds红色
# 这里是修改标签
# We want to show all ticks...
ax.set_xticks(np.arange(len(x)))
ax.set_yticks(np.arange(len(y)))
# ... and label theym with the respective list entries
ax.set_xticklabels(x)
ax.set_yticklabels(y)
# 因为x轴的标签太长了,需要旋转一下,更加好看
# Rotate the tick labels and set their alignment.
plt.setp(ax.get_xticklabels(), rotation=0, ha="right",
rotation_mode="anchor")
# 添加每个热力块的具体数值
# Loop over data dimensions and create text annotations.
for i in range(len(y)):
for j in range(len(x)):
text = ax.text(j, i, height[i, j],
ha="center", va="center", color="black", size="7")
ax.set_title("Block " + name.split(".")[0])
fig.tight_layout()
plt.colorbar(im)
plt.savefig(name)
plt.show()
#保存为csv
def save_csv(path,numpy_array):
dataframe = pd.DataFrame(numpy_array)
dataframe.to_csv(path, header=False, index=False, sep=',')
if __name__ == '__main__':
pcd_read = o3d.io.read_point_cloud("819-1 - start-cut-pre.pcd")
# 记录时间
old_time = time.time()
# pcd_read = o3d.io.read_point_cloud("data/819-1 - start-cut-pre.pcd")
len_xy = (8, 0.8, 0.3) # 划分小方格--自定义--y长0.8,x划分8个,中间水沟0.3
pcd_zero = zero(pcd_read)
del pcd_read
gc.collect()
height,inclination = cal_para(pcd_zero, len_xy)
# print("宽x: {0}, 长y: {1} ,总体平均: {2}".format(avg_height_x(height), avg_height_y(height), avg_height(height)))
height_rev = reverse_matrix(height)
inclination_rev = reverse_matrix(inclination)
# plot_matrix(height_rev, "result2/height.png")
np.savetxt("result2/height.csv", height_rev)
np.savetxt("result2/inclination.csv", inclination_rev)
# save_csv(height,"result/height.csv")
current_time = time.time()
print("运行时间为" + str(current_time - old_time) + "s")
# label = ["{}".format(i) for i in range(0, height.shape[0])]
# plot_confusion_matrix(height, label)
# print("x y z : ", np.asarray(pcd_zero.points).max(0))
# print("min : {0}".format(np.max(np_points, 0)))