我有一些测试代码使用rsgislib来分割图像。从rsgislib的文档中我运行:
from rsgislib.segmentation import segutils
inputImg = 'clipped_geotiff_image.tif'
outputClumps = 'clipped_jers1palsar_stack_clumps_elim_final.tif'
outputMeanImg = 'clipped_jers1palsar_stack_clumps_elim_final_mean.tif'
segutils.runShepherdSegmentation(inputImg, outputClumps, outputMeanImg, minPxls=100)
在尝试通过以下方式显式添加后,outputMeanImg不包含坐标系和其他空间信息:
rsgislib.imageutils.copySpatialAndProjFromImage(outputMeanImg, inputImg)
或
rsgislib.imageutils.assignProj(outputMeanImg, rsgislib.imageutils.getWKTProjFromImage(outputMeanImg), None)
有关详细信息,我在下面提供了gdalinfo的输出。任何人都可以建议如何设置corrdinate系统和其他投影信息或将子数据集BAND1 / DATA转换为传统的光栅文件?
坦克提前。
=====
%gdalinfo clip_jers1palsar_stack_clumps_elim_final_mean.kea
驱动程序:HDF5 /分层数据格式版本5文件: clip_jers1palsar_stack_clumps_elim_final_mean.kea尺寸为512,512 坐标系是''元数据:BAND1_DATA_BLOCK_SIZE = 256d
BAND1_DATA_CLASS = IMAGE BAND1_DATA_IMAGE_VERSION = 1.2
BAND1_NO_DATA_VAL_NO_DATA_DEFINED =
BAND1_OVERVIEWS_OVERVIEW1_BLOCK_SIZE = 256D
BAND1_OVERVIEWS_OVERVIEW1_CLASS = IMAGE
BAND1_OVERVIEWS_OVERVIEW1_IMAGE_VERSION = 1.2
BAND1_OVERVIEWS_OVERVIEW2_BLOCK_SIZE = 187D
BAND1_OVERVIEWS_OVERVIEW2_CLASS = IMAGE
BAND1_OVERVIEWS_OVERVIEW2_IMAGE_VERSION = 1.2
BAND1_OVERVIEWS_OVERVIEW3_BLOCK_SIZE = 93D
BAND1_OVERVIEWS_OVERVIEW3_CLASS = IMAGE
BAND1_OVERVIEWS_OVERVIEW3_IMAGE_VERSION = 1.2
BAND1_OVERVIEWS_OVERVIEW4_BLOCK_SIZE = 46D
BAND1_OVERVIEWS_OVERVIEW4_CLASS = IMAGE
BAND1_OVERVIEWS_OVERVIEW4_IMAGE_VERSION = 1.2 Subdatasets:
SUBDATASET_1_NAME = HDF5: “clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 / ATT / DATA / FLOAT SUBDATASET_1_DESC = [256x1] // BAND1 / ATT / DATA / FLOAT(64位) 浮点)
SUBDATASET_2_NAME = HDF5:“clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 / DATA SUBDATASET_2_DESC = [2500x1500] // BAND1 / DATA(16位无符号整数)
SUBDATASET_3_NAME = HDF5: “clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 /概述/ OVERVIEW1 SUBDATASET_3_DESC = [625x375] // BAND1 / OVERVIEWS / OVERVIEW1(16位 无符号整数)
SUBDATASET_4_NAME = HDF5: “clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 /概述/ OVERVIEW2 SUBDATASET_4_DESC = [312x187] // BAND1 / OVERVIEWS / OVERVIEW2(16位 无符号整数)
SUBDATASET_5_NAME = HDF5: “clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 /概述/ OVERVIEW3 SUBDATASET_5_DESC = [156x93] // BAND1 / OVERVIEWS / OVERVIEW3(16位 无符号整数)
SUBDATASET_6_NAME = HDF5: “clip_jers1palsar_stack_clumps_elim_final_mean.kea”:// BAND1 /概述/ OVERVIEW4 SUBDATASET_6_DESC = [78x46] // BAND1 / OVERVIEWS / OVERVIEW4(16位无符号) 整数)角坐标:左上角(0.0,0.0)左下角 (0.0,512.0)右上(512.0,0.0)右下(512.0, 512.0)中心(256.0,256.0)
答案 0 :(得分:0)
我找到了部分解决方案,但没有找到rsgislib。
使用GDAL,我可以打开一个子数据集:
from osgeo import gdal
# get the projection and geotransform from the original input image
ids = gdal.Open(inputImg, gdal.GA_ReadOnly)
gt = ids.GetGeoTransform()
wkt = ids.GetProjectionRef()
print gt
del ids
# now set the projection and geotransform for the specific subdatsert
infile = 'HDF5:"'+outputMeanImg+'"://BAND1/DATA'
ods = gdal.Open(infile, gdal.GA_ReadOnly)
ods.SetProjection(wkt)
ods.SetGeoTransform(gt)
del ods
上面的代码设置了我感兴趣的子数据集的投影和地理转换,但请注意我打开输出为ReadOnly。如果我将其设置为更新,则会失败,并且它实际上会更新所选数据集,即使它是以只读模式打开的。我认为这是一个错误,但它似乎通过创建一个新的.aux.xml文件来工作。
可能有更简单/更好的方法来做到这一点,但现在这种方法有效。