准备
安装python以及gdal
馨意:基于python的遥感图像处理(1.1)--Anaconda安装步骤zhuanlan.zhihu.com正文
本文以提取MODIS植被指数产品MOD13A3的NDVI数据为例:
首先打开hdf数据集并查看子数据集和元数据:
# gdal打开hdf数据集
datasets = gdal.Open(r"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf")
# 获取hdf中的子数据集
SubDatasets = datasets.GetSubDatasets()
# 获取子数据集的个数
SubDatasetsNum = len(datasets.GetSubDatasets())
# 输出各子数据集的信息
print("子数据集一共有{0}个: ".format(SubDatasetsNum))
for i in range(SubDatasetsNum):
print(datasets.GetSubDatasets()[i])
# 获取hdf中的元数据
Metadata = datasets.GetMetadata()
# 获取元数据的个数
MetadataNum = len(Metadata)
# 输出各子数据集的信息
print("元数据一共有{0}个: ".format(MetadataNum))
for key,value in Metadata.items():
print('{key}:{value}'.format(key = key, value = value))
输出为:
子数据集一共有11个:
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly NDVI"', '[1200x1200] 1 km monthly NDVI MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly EVI"', '[1200x1200] 1 km monthly EVI MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly VI Quality"', '[1200x1200] 1 km monthly VI Quality MOD_Grid_monthly_1km_VI (16-bit unsigned integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly red reflectance"', '[1200x1200] 1 km monthly red reflectance MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly NIR reflectance"', '[1200x1200] 1 km monthly NIR reflectance MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly blue reflectance"', '[1200x1200] 1 km monthly blue reflectance MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly MIR reflectance"', '[1200x1200] 1 km monthly MIR reflectance MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly view zenith angle"', '[1200x1200] 1 km monthly view zenith angle MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly sun zenith angle"', '[1200x1200] 1 km monthly sun zenith angle MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly relative azimuth angle"', '[1200x1200] 1 km monthly relative azimuth angle MOD_Grid_monthly_1km_VI (16-bit integer)')
('HDF4_EOS:EOS_GRID:"E:Remote_Sensing_DataTVDIMOD13A32012MOD13A3.A2012001.h26v04.005.hdf":MOD_Grid_monthly_1km_VI:"1 km monthly pixel raliability"', '[1200x1200] 1 km monthly pixel raliability MOD_Grid_monthly_1km_VI (8-bit integer)')
元数据一共有163个:
ALGORITHMPACKAGEACCEPTANCEDATE:102004
ALGORITHMPACKAGEMATURITYCODE:Normal
......元数据太多,只展示这几个为例
DATACOLUMNS:1200
DATAROWS:1200
......
GRINGPOINTLATITUDE.1:39.8624390029821, 50.1154198026055, 49.9369825110409, 39.7325278324118
GRINGPOINTLONGITUDE.1:104.235558676363, 124.607874257256, 140.825259076411, 117.770602913265
......
VERTICALTILENUMBER:04
WESTBOUNDINGCOORDINATE:104.43258313171
我们可以发现子数据集中的第一个数据为NDVI数据,我们将数值提取出来:
# 第一个子数据集合,也就是NDVI数据
DatasetNDVI = datasets.GetSubDatasets()[0][0]
RasterNDVI = gdal.Open(DatasetNDVI)
NDVI = RasterNDVI.ReadAsArray()
如果我们转成tif格式数据,需要知道图像的仿射变换矩阵,仿射变换矩阵可以通过图像的四个角的经纬度计算得出(下面链接可以帮助了解一下仿射变换矩阵):
馨意:python+GDAL遥感图像像素坐标与地理坐标转换(仿射变换)zhuanlan.zhihu.com# 获取四个角的维度
Latitudes = Metadata["GRINGPOINTLATITUDE.1"]
# 采用", "进行分割
LatitudesList = Latitudes.split(", ")
# 获取四个角的经度
Longitude = Metadata["GRINGPOINTLONGITUDE.1"]
# 采用", "进行分割
LongitudeList = Longitude.split(", ")
# 图像四个角的地理坐标
GeoCoordinates = np.zeros((4, 2), dtype = "float32")
GeoCoordinates[0] = np.array([float(LongitudeList[0]),float(LatitudesList[0])])
GeoCoordinates[1] = np.array([float(LongitudeList[1]),float(LatitudesList[1])])
GeoCoordinates[2] = np.array([float(LongitudeList[2]),float(LatitudesList[2])])
GeoCoordinates[3] = np.array([float(LongitudeList[3]),float(LatitudesList[3])])
# 列数
Columns = float(Metadata["DATACOLUMNS"])
# 行数
Rows = float(Metadata["DATAROWS"])
# 图像四个角的图像坐标
PixelCoordinates = np.array([[0, 0],
[Columns - 1, 0],
[Columns - 1, Rows - 1],
[0, Rows - 1]], dtype = "float32")
# 计算仿射变换矩阵
from scipy.optimize import leastsq
def func(i):
Transform0, Transform1, Transform2, Transform3, Transform4, Transform5 = i[0], i[1], i[2], i[3], i[4], i[5]
return [Transform0 + PixelCoordinates[0][0] * Transform1 + PixelCoordinates[0][1] * Transform2 - GeoCoordinates[0][0],
Transform3 + PixelCoordinates[0][0] * Transform4 + PixelCoordinates[0][1] * Transform5 - GeoCoordinates[0][1],
Transform0 + PixelCoordinates[1][0] * Transform1 + PixelCoordinates[1][1] * Transform2 - GeoCoordinates[1][0],
Transform3 + PixelCoordinates[1][0] * Transform4 + PixelCoordinates[1][1] * Transform5 - GeoCoordinates[1][1],
Transform0 + PixelCoordinates[2][0] * Transform1 + PixelCoordinates[2][1] * Transform2 - GeoCoordinates[2][0],
Transform3 + PixelCoordinates[2][0] * Transform4 + PixelCoordinates[2][1] * Transform5 - GeoCoordinates[2][1],
Transform0 + PixelCoordinates[3][0] * Transform1 + PixelCoordinates[3][1] * Transform2 - GeoCoordinates[3][0],
Transform3 + PixelCoordinates[3][0] * Transform4 + PixelCoordinates[3][1] * Transform5 - GeoCoordinates[3][1]]
# 最小二乘法求解
GeoTransform = leastsq(func,np.asarray((1,1,1,1,1,1)))
print(GeoTransform)
为了方便处理数据,我们获取数据的时间并以此为文件名保存:
# 获取数据时间
date = Metadata["RANGEBEGINNINGDATE"]
# 保存为tif
def array2raster(TifName, GeoTransform, array):
cols = array.shape[1] # 矩阵列数
rows = array.shape[0] # 矩阵行数
driver = gdal.GetDriverByName('GTiff')
outRaster = driver.Create(TifName, cols, rows, 1, gdal.GDT_Float32)
# 括号中两个0表示起始像元的行列号从(0,0)开始
outRaster.SetGeoTransform(tuple(GeoTransform))
# 获取数据集第一个波段,是从1开始,不是从0开始
outband = outRaster.GetRasterBand(1)
outband.WriteArray(array)
outRasterSRS = osr.SpatialReference()
# 代码4326表示WGS84坐标
outRasterSRS.ImportFromEPSG(4326)
outRaster.SetProjection(outRasterSRS.ExportToWkt())
outband.FlushCache()
TifName = date + ".tif"
array2raster(TifName, GeoTransform[0], NDVI)
至此,大功告成。
批量转换代码:
import gdal, osr
import numpy as np
import os
# 数组保存为tif
def array2raster(TifName, GeoTransform, array):
cols = array.shape[1] # 矩阵列数
rows = array.shape[0] # 矩阵行数
driver = gdal.GetDriverByName('GTiff')
outRaster = driver.Create(TifName, cols, rows, 1, gdal.GDT_Float32)
# 括号中两个0表示起始像元的行列号从(0,0)开始
outRaster.SetGeoTransform(tuple(GeoTransform))
# 获取数据集第一个波段,是从1开始,不是从0开始
outband = outRaster.GetRasterBand(1)
outband.WriteArray(array)
outRasterSRS = osr.SpatialReference()
# 代码4326表示WGS84坐标
outRasterSRS.ImportFromEPSG(4326)
outRaster.SetProjection(outRasterSRS.ExportToWkt())
outband.FlushCache()
# hdf批量转tif
def hdf2tif_batch(hdfFolder):
# 获取文件夹内的文件名
hdfNameList = os.listdir(hdfFolder)
for i in range(len(hdfNameList)):
# 判断当前文件是否为HDF文件
if(os.path.splitext(hdfNameList[i])[1] == ".hdf"):
hdfPath = hdfFolder+"/"+hdfNameList[i]
# gdal打开hdf数据集
datasets = gdal.Open(hdfPath)
# 获取hdf中的元数据
Metadata = datasets.GetMetadata()
# 获取四个角的维度
Latitudes = Metadata["GRINGPOINTLATITUDE.1"]
# 采用", "进行分割
LatitudesList = Latitudes.split(", ")
# 获取四个角的经度
Longitude = Metadata["GRINGPOINTLONGITUDE.1"]
# 采用", "进行分割
LongitudeList = Longitude.split(", ")
# 图像四个角的地理坐标
GeoCoordinates = np.zeros((4, 2), dtype = "float32")
GeoCoordinates[0] = np.array([float(LongitudeList[0]),float(LatitudesList[0])])
GeoCoordinates[1] = np.array([float(LongitudeList[1]),float(LatitudesList[1])])
GeoCoordinates[2] = np.array([float(LongitudeList[2]),float(LatitudesList[2])])
GeoCoordinates[3] = np.array([float(LongitudeList[3]),float(LatitudesList[3])])
# 列数
Columns = float(Metadata["DATACOLUMNS"])
# 行数
Rows = float(Metadata["DATAROWS"])
# 图像四个角的图像坐标
PixelCoordinates = np.array([[0, 0],
[Columns - 1, 0],
[Columns - 1, Rows - 1],
[0, Rows - 1]], dtype = "float32")
# 计算仿射变换矩阵
from scipy.optimize import leastsq
def func(i):
Transform0, Transform1, Transform2, Transform3, Transform4, Transform5 = i[0], i[1], i[2], i[3], i[4], i[5]
return [Transform0 + PixelCoordinates[0][0] * Transform1 + PixelCoordinates[0][1] * Transform2 - GeoCoordinates[0][0],
Transform3 + PixelCoordinates[0][0] * Transform4 + PixelCoordinates[0][1] * Transform5 - GeoCoordinates[0][1],
Transform0 + PixelCoordinates[1][0] * Transform1 + PixelCoordinates[1][1] * Transform2 - GeoCoordinates[1][0],
Transform3 + PixelCoordinates[1][0] * Transform4 + PixelCoordinates[1][1] * Transform5 - GeoCoordinates[1][1],
Transform0 + PixelCoordinates[2][0] * Transform1 + PixelCoordinates[2][1] * Transform2 - GeoCoordinates[2][0],
Transform3 + PixelCoordinates[2][0] * Transform4 + PixelCoordinates[2][1] * Transform5 - GeoCoordinates[2][1],
Transform0 + PixelCoordinates[3][0] * Transform1 + PixelCoordinates[3][1] * Transform2 - GeoCoordinates[3][0],
Transform3 + PixelCoordinates[3][0] * Transform4 + PixelCoordinates[3][1] * Transform5 - GeoCoordinates[3][1]]
# 最小二乘法求解
GeoTransform = leastsq(func,np.asarray((1,1,1,1,1,1)))
# 获取数据时间
date = Metadata["RANGEBEGINNINGDATE"]
# 第一个子数据集合,也就是NDVI数据
DatasetNDVI = datasets.GetSubDatasets()[0][0]
RasterNDVI = gdal.Open(DatasetNDVI)
NDVI = RasterNDVI.ReadAsArray()
TifName = date + ".tif"
array2raster(TifName, GeoTransform[0], NDVI)
print(TifName,"Saved successfully!")
hdf2tif_batch(r"E:Remote_Sensing_DataTVDIMOD13A32012")
输出:
2012-01-01.tif Saved successfully!
2012-02-01.tif Saved successfully!
2012-03-01.tif Saved successfully!
2012-04-01.tif Saved successfully!
2012-05-01.tif Saved successfully!
2012-06-01.tif Saved successfully!
2012-07-01.tif Saved successfully!
2012-08-01.tif Saved successfully!
2012-09-01.tif Saved successfully!
2012-10-01.tif Saved successfully!
2012-11-01.tif Saved successfully!
2012-12-01.tif Saved successfully!