Tiff 图像到 numpy 数组

Question

import rasterio as rio
from rasterio.plot import show
from sklearn import cluster
import matplotlib.pyplot as plt
import numpy as np
import glob

for filepath in glob.iglob('./dengue3/*.tiff'):
    elhas_raster = rio.open(filepath)
    elhas_arr = elhas_raster.read() # read the opened image
    vmin, vmax = np.nanpercentile(elhas_arr, (5,95))  # 5-95% contrast stretch
    # create an empty array with same dimension and data type
    imgxyb = np.empty((elhas_raster.height, elhas_raster.width, elhas_raster.count), elhas_raster.meta['dtype'])
    # loop through the raster's bands to fill the empty array
    for band in range(imgxyb.shape[2]):
        imgxyb[:,:,band] = elhas_raster.read(band+1)
    #print(imgxyb.shape)
    # convert to 1d array
    img1d=imgxyb[:,:,:7].reshape((imgxyb.shape[0]*imgxyb.shape[1],imgxyb.shape[2]))
    #print(img1d.shape)

上面的代码我用来读取文件夹中的 tiff 图像并获取数组。但是，输出是 -

ValueError: cannot reshape array of size 6452775 into shape (921825,12)

图像为 12 波段。我尝试在上面的代码中使用 12 代替 7，但代码没有执行。我该如何解决这个问题？感谢您的宝贵时间。

Answer 1

你改变了你试图重塑的索引的大小，但没有改变 reshape 命令参数：

img1d=imgxyb[:,:,:7].reshape((imgxyb.shape[0]*imgxyb.shape[1],imgxyb.shape[2]))

这应该是：

img1d=imgxyb[:,:,:7].reshape((imgxyb.shape[0]*imgxyb.shape[1],7))