遥感图像云检测数据集CHLandsat8

CHLandsat8数据集

数据集存储地址：https://pan.baidu.com/s/1dwKOuq-kDVAXebV8jy-DaQ?pwd=91jn
裁剪和拼接程序：https://github.com/HaiLei-Fly/CHLandsat8

CHLandsat8 数据集： 本文创建的一个中国Landsat 8 高分辨率云检测数据集，其中包含了 2021年 1 月至 2021 年 12 月 Landsat 8 卫星从中国不同地区采集的 64 个全场景。据我们所知，CHLandsat8 可能是从 Landsat 8 卫星最早收集中国地区的遥感图像云检测数据集。而且与以往的数据集相比，CHLandsat8 中的场景更加复杂多样，实现高精度的云检测更具挑战性。 该数据集覆盖中国西北地区、 北方地区、 青藏地区和南方地区， 包含各种土地覆盖类
型，包括城市、冰雪、草原、山川、森林、海洋和沙漠。数据集是自然彩色图像，尺寸大约为 8000*8000像素。此外，数据集的参考云 mask 已进行标注，并可在网上查阅。我们相信，开放的 CHLandsat8 数据集有助于促进云检测的研究。
在建立数据集的过程中，专家手动逐个像素地标记云在图像中的位置，并通过分别用 1 和 0 标记云和背景的像素值来创建参考 mask。为了确保标签的准确性，参考 mask 已经过迭代检查和校正。在实验中，本文从数据集中随机选取 44 幅图像作为训练集 CHLandsat8-TR ， 20 幅 图 像 作 为 测 试 集CHLandsat8-TE。
由于 GPU 的内存有限，数据集不同场景图像被裁剪成 352×352 大小，详细数据如表所示。

图像裁剪和拼接

图像裁剪程序：

# -*- coding:utf-8 -*-
import os
import matplotlib.pyplot as plt
import cv2
import numpy as np
import math
import glob

"""
输入：图片路径(path+filename)，裁剪获得小图片的列数、行数（也即宽、高）
输出：无
"""
def crop_one_picture(path,filename,cols,rows):
    img=cv2.imread(path+filename,1)##读取彩色图像，图像的透明度(alpha通道)被忽略，默认参数;灰度图像;读取原始图像，包括alpha通道;可以用1，0，-1来表示
    # img=cv2.imread(path+filename,-1) # mask
    # img= cv2.copyMakeBorder(img, top=97, bottom=98, left=152, right=153,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 1
    # img= cv2.copyMakeBorder(img, top=87, bottom=88, left=147, right=148,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 2
    # img= cv2.copyMakeBorder(img, top=137, bottom=138, left=36, right=37,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 3
    # img= cv2.copyMakeBorder(img, top=132, bottom=133, left=31, right=32,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 4
    # img= cv2.copyMakeBorder(img, top=12, bottom=13, left=62, right=63,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 5
    # img= cv2.copyMakeBorder(img, top=107, bottom=108, left=6, right=7,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 6
    # img= cv2.copyMakeBorder(img, top=107, bottom=108, left=167, right=168,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 7
    # img= cv2.copyMakeBorder(img, top=97, bottom=98, left=162, right=163,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 8
    # img= cv2.copyMakeBorder(img, top=82, bottom=83, left=152, right=153,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 9
    # img= cv2.copyMakeBorder(img, top=57, bottom=58, left=122, right=123,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 10
    # img= cv2.copyMakeBorder(img, top=6, bottom=7, left=81, right=82,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 11
    # img= cv2.copyMakeBorder(img, top=62, bottom=63, left=112, right=113,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 12
    # img= cv2.copyMakeBorder(img, top=22, bottom=23, left=67, right=68,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 13
    # img= cv2.copyMakeBorder(img, top=127, bottom=128, left=16, right=17,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 14
    # img= cv2.copyMakeBorder(img, top=21, bottom=22, left=101, right=102,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 15
    # img= cv2.copyMakeBorder(img, top=158, bottom=159, left=27, right=28,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 16
    # img= cv2.copyMakeBorder(img, top=31, bottom=32, left=106, right=107,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 17
    # img= cv2.copyMakeBorder(img, top=97, bottom=98, left=152, right=153,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 18
    # img= cv2.copyMakeBorder(img, top=152, bottom=153, left=51, right=52,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 19
    # img= cv2.copyMakeBorder(img, top=142, bottom=143, left=51, right=52,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 20
    img= cv2.copyMakeBorder(img, top=167, bottom=168, left=61, right=62,
                        borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 21
    # img= cv2.copyMakeBorder(img, top=132, bottom=133, left=31, right=32,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 22
    # img= cv2.copyMakeBorder(img, top=152, bottom=153, left=41, right=42,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 23
    # img= cv2.copyMakeBorder(img, top=82, bottom=83, left=147, right=148,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 24
    # img= cv2.copyMakeBorder(img, top=16, bottom=17, left=91, right=92,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 25
    # img= cv2.copyMakeBorder(img, top=57, bottom=58, left=122, right=123,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 26
    # img= cv2.copyMakeBorder(img, top=82, bottom=83, left=147, right=148,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 27
    # img= cv2.copyMakeBorder(img, top=52, bottom=53, left=117, right=118,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 28
    # img= cv2.copyMakeBorder(img, top=137, bottom=138, left=36, right=37,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 29
    # img= cv2.copyMakeBorder(img, top=132, bottom=133, left=26, right=27,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 30
    # img= cv2.copyMakeBorder(img, top=47, bottom=48, left=102, right=103,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 31
    # img= cv2.copyMakeBorder(img, top=172, bottom=173, left=66, right=67,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 32
    # img= cv2.copyMakeBorder(img, top=1, bottom=2, left=61, right=62,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 33
    # img= cv2.copyMakeBorder(img, top=16, bottom=17, left=91, right=92,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 34
    # img= cv2.copyMakeBorder(img, top=152, bottom=153, left=51, right=52,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 35
    # img= cv2.copyMakeBorder(img, top=107, bottom=108, left=172, right=173,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 36
    # img= cv2.copyMakeBorder(img, top=112, bottom=113, left=167, right=168,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 37
    # img= cv2.copyMakeBorder(img, top=92, bottom=93, left=147, right=148,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 38
    # img= cv2.copyMakeBorder(img, top=87, bottom=88, left=147, right=148,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 39
    # img= cv2.copyMakeBorder(img, top=12, bottom=13, left=72, right=73,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 40
    # img= cv2.copyMakeBorder(img, top=127, bottom=128, left=11, right=12,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 41
    # img= cv2.copyMakeBorder(img, top=112, bottom=113, left=1, right=2,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 42
    # img= cv2.copyMakeBorder(img, top=11, bottom=12, left=76, right=77,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 43
    # img= cv2.copyMakeBorder(img, top=157, bottom=158, left=51, right=52,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 44
    ####### Test
    # img= cv2.copyMakeBorder(img, top=142, bottom=143, left=46, right=47,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 1
    # img= cv2.copyMakeBorder(img, top=67, bottom=68, left=127, right=128,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 2
    # img= cv2.copyMakeBorder(img, top=132, bottom=133, left=36, right=37,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 3
    # img= cv2.copyMakeBorder(img, top=127, bottom=128, left=16, right=17,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 4
    # img= cv2.copyMakeBorder(img, top=21, bottom=22, left=106, right=107,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 5
    # img= cv2.copyMakeBorder(img, top=127, bottom=128, left=11, right=12,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 6
    # img= cv2.copyMakeBorder(img, top=102, bottom=103, left=167, right=168,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 7
    # img= cv2.copyMakeBorder(img, top=97, bottom=98, left=162, right=163,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 8
    # img= cv2.copyMakeBorder(img, top=132, bottom=133, left=16, right=17,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 9
    # img= cv2.copyMakeBorder(img, top=107, bottom=108, left=1, right=2,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 10
    # img= cv2.copyMakeBorder(img, top=31, bottom=32, left=111, right=112,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 11
    # img= cv2.copyMakeBorder(img, top=107, bottom=108, left=1, right=2,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 12
    # img= cv2.copyMakeBorder(img, top=21, bottom=22, left=96, right=97,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 13
    # img= cv2.copyMakeBorder(img, top=6, bottom=7, left=76, right=77,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 14
    # img= cv2.copyMakeBorder(img, top=152, bottom=153, left=46, right=47,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 15
    # img= cv2.copyMakeBorder(img, top=92, bottom=93, left=152, right=153,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 16
    # img= cv2.copyMakeBorder(img, top=6, bottom=7, left=81, right=82,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 17
    # img= cv2.copyMakeBorder(img, top=162, bottom=163, left=51, right=52,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 18
    # img= cv2.copyMakeBorder(img, top=162, bottom=163, left=61, right=62,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 19
    # img= cv2.copyMakeBorder(img, top=1, bottom=2, left=66, right=67,
    #                     borderType=cv2.BORDER_CONSTANT, value=[0, 0, 0]) # 20
    
    sum_rows=img.shape[0]   #高度
    sum_cols=img.shape[1]    #宽度
    save_path=path+"\\crop{0}_{1}\\".format(cols,rows)  #保存的路径
    if not os.path.exists(save_path):
        os.makedirs(save_path)
    print("裁剪所得{0}列图片，{1}行图片.".format(int(sum_cols/cols),int(sum_rows/rows))) 
    k = 0
    for i in range(int(sum_rows/rows)):
        for j in range(int(sum_cols/cols)):
            k += 1
            # cv2.imwrite(save_path+'patch_{}_{}_by_{}_'.format(k, i+1, j+1)+ filename.split('.')[0] + os.path.splitext(filename)[1],img[i*cols:(i+1)*cols,j*rows:(j+1)*rows,:]) # L image
            cv2.imwrite(save_path+'patch_{}_{}_by_{}_'.format(k, i+1, j+1)+ filename.split('.')[0] + os.path.splitext(filename)[1],img[i*cols:(i+1)*cols,j*rows:(j+1)*rows]) # mask
    print("裁剪完成，得到{0}张图片.".format(int(sum_cols/cols)*int(sum_rows/rows)))
    print("文件保存在{0}".format(save_path))
    
if __name__ == '__main__':

    path='G:\\CHLandsat8\\LandsatNC\\Train\\patch\\'   #要裁剪的图片所在的文件夹
    picture_names = sorted(glob.glob(path  + '*.png'))
    num = 0
    for picture_name in picture_names:
        print(picture_name)
        name = picture_name.split('\\')[5] 
        print(name)
        crop_one_picture(path,name,352,352)
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

图像拼接程序：

from cv2 import cv2
import numpy as np
import os
# 分割后的图片的文件夹，以及拼接后要保存的文件夹
pic_path = 'H:/FAANet/Pre/CHLandsat8/joint/'
pic_target = 'H:/FAANet/Pre/38_Cloud_Train/CHLandsat8-joint/'
# 数组保存分割后图片的列数和行数，注意分割后图片的格式为x_x.jpg，x从1开始
num_width_list = []
num_lenght_list = []
# 读取文件夹下所有图片的名称
picture_names = os.listdir(pic_path)
if len(picture_names)==0:
    print("没有文件")
else:
    # 获取分割后图片的尺寸
    img_1_1 = cv2.imread(pic_path + 'patch_1_1_by_1_LC08_L1TP_117040_20210625_20210630_01_T1.png')
    (width, length, depth) = img_1_1.shape
    # 分割名字获得行数和列数，通过数组保存分割后图片的列数和行数
    for picture_name in picture_names:
        num_width_list.append(int(picture_name.split("_")[-10]))
        num_lenght_list.append(int((picture_name.split("_")[-8])))
    # 取其中的最大值
    num_width = max(num_width_list)
    num_length = max(num_lenght_list)
    # 预生成拼接后的图片
    splicing_pic = np.zeros((num_width*width, num_length*length, depth))
    # 循环复制
    for idx in range(0, 1):
        k = 0
        splicing_pic = np.zeros((num_width*width, num_length*length, depth))
        for i in range(1, num_width+1):
            for j in range(1, num_length+1):
                    k += 1
                    img_part = cv2.imread(pic_path + 'patch_{}_{}_by_{}_LC08_L1TP_117040_20210625_20210630_01_T1.png'.format(k, i, j),1)         
                    splicing_pic[ width*(i-1) : width*i, length*(j-1) : length*j, :] = img_part
                    print(splicing_pic.shape)
        cv2.imwrite(pic_target + 'LC08_L1TP_' + picture_names[idx].split("_")[-5] + '_' +  picture_names[idx].split("_")[-4] + '_'  +  picture_names[idx].split("_")[-3]  + '_01_T1.png', splicing_pic)
    print("done!!!")

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

数据集图像示例

彩色图像示例

彩色图像裁剪示例

GT图像示例

GT图像裁剪示例

希望本文对大家有帮助，上文若有不妥之处，欢迎指正

分享决定高度，学习拉开差距