RPC | 高分三号雷达数据RPC校正代码

RPC（rational polynomial coefficients），有理多项式系数。RPC是传感器严格几何模型的拟合形式。是根据卫星平台载荷测量的平台运行轨迹参数、姿态参数、传感器安装参数及传感器内部几何参数构建的像-地关系几何模型，一共有90个参数。

相关理论不再展示，感兴趣去网上搜，现在是基于高分三号进行RPC校正，代码如下：

from osgeo import gdal
import re
import os

#这是一个读取RPC文件的函数
def read_rpb (rpbfile):
    with open(rpbfile, 'r') as f:
        buff = f.read()
        # 关键字1,2(修改引号间的内容)
        # 命名参考的网址：http://geotiff.maptools.org/rpc_prop.html
        ERR_BIAS1 = 'errBias      = '  # 错误-偏差。图像中所有点的RMS偏差误差（单位：米/水平轴）（-1.0，如果未知）
        ERR_BIAS2 = ';'

        ERR_RAND1 = 'errRand      = '  # 错误-随机。图像中每个点的每个水平轴的RMS随机误差，以米为单位（如果未知，则为-1.0）
        ERR_RAND2 = ';'

        LINE_OFF1 = 'lineOffset   = '  # 线偏移
        LINE_OFF2 = ';'

        SAMP_OFF1 = 'sampOffset   = '  # 样品偏移
        SAMP_OFF2 = ';'

        LAT_OFF1 = 'latOffset    =  '  # 大地纬度偏移
        LAT_OFF2 = ';'

        LONG_OFF1 = 'longOffset   = '  # 大地经度偏移
        LONG_OFF2 = ';'

        HEIGHT_OFF1 = 'heightOffset =   '  # 大地高程偏移
        HEIGHT_OFF2 = ';'

        LINE_SCALE1 = 'lineScale    =  '  # 线刻度
        LINE_SCALE2 = ';'

        SAMP_SCALE1 = 'sampScale    =  '  # 样本量表
        SAMP_SCALE2 = ';'

        LAT_SCALE1 = 'latScale     =    '  # 大地纬度刻度
        LAT_SCALE2 = ';'

        LONG_SCALE1 = 'longScale    =   '  # 大地经度标度
        LONG_SCALE2 = ';'

        HEIGHT_SCALE1 = 'heightScale  = '  # 大地高度标尺
        HEIGHT_SCALE2 = ';'

        LINE_NUM_COEFF1 = 'lineNumCoef  = ('  # 线分子系数。r n方程分子中的多项式的二十个系数。
        LINE_NUM_COEFF2 = ');'

        LINE_DEN_COEFF1 = 'lineDenCoef = ('  # 线分母系数。r n方程的分母中的多项式的二十个系数。
        LINE_DEN_COEFF2 = ');'

        SAMP_NUM_COEFF1 = 'sampNumCoef = ('  # 样本分子系数。c n方程分子中的多项式的二十个系数。
        SAMP_NUM_COEFF2 = ');'

        SAMP_DEN_COEFF1 = 'sampDenCoef = ('  # 样本分母系数。c n方程的分母中的多项式的二十个系数。
        SAMP_DEN_COEFF2 = ');'

        # 正则化提取数值
        pat_ERR_BIAS = re.compile(ERR_BIAS1 + '(.*?)' + ERR_BIAS2, re.S)
        result_ERR_BIAS = pat_ERR_BIAS.findall(buff)
        ERR_BIAS = result_ERR_BIAS[0]

        pat_ERR_RAND = re.compile(ERR_RAND1 + '(.*?)' + ERR_RAND2, re.S)
        result_ERR_RAND = pat_ERR_RAND.findall(buff)
        ERR_RAND = result_ERR_RAND[0]

        pat_LINE_OFF = re.compile(LINE_OFF1 + '(.*?)' + LINE_OFF2, re.S)
        result_LINE_OFF = pat_LINE_OFF.findall(buff)
        LINE_OFF = result_LINE_OFF[0]

        pat_SAMP_OFF = re.compile(SAMP_OFF1 + '(.*?)' + SAMP_OFF2, re.S)
        result_SAMP_OFF = pat_SAMP_OFF.findall(buff)
        SAMP_OFF = result_SAMP_OFF[0]

        pat_LAT_OFF = re.compile(LAT_OFF1 + '(.*?)' + LAT_OFF2, re.S)
        result_LAT_OFF = pat_LAT_OFF.findall(buff)
        LAT_OFF = result_LAT_OFF[0]

        pat_LONG_OFF = re.compile(LONG_OFF1 + '(.*?)' + LONG_OFF2, re.S)
        result_LONG_OFF = pat_LONG_OFF.findall(buff)
        LONG_OFF = result_LONG_OFF[0]

        pat_HEIGHT_OFF = re.compile(HEIGHT_OFF1 + '(.*?)' + HEIGHT_OFF2, re.S)
        result_HEIGHT_OFF = pat_HEIGHT_OFF.findall(buff)
        HEIGHT_OFF = result_HEIGHT_OFF[0]

        pat_LINE_SCALE = re.compile(LINE_SCALE1 + '(.*?)' + LINE_SCALE2, re.S)
        result_LINE_SCALE = pat_LINE_SCALE.findall(buff)
        LINE_SCALE = result_LINE_SCALE[0]

        pat_SAMP_SCALE = re.compile(SAMP_SCALE1 + '(.*?)' + SAMP_SCALE2, re.S)
        result_SAMP_SCALE = pat_SAMP_SCALE.findall(buff)
        SAMP_SCALE = result_SAMP_SCALE[0]

        pat_LAT_SCALE = re.compile(LAT_SCALE1 + '(.*?)' + LAT_SCALE2, re.S)
        result_LAT_SCALE = pat_LAT_SCALE.findall(buff)
        LAT_SCALE = result_LAT_SCALE[0]

        pat_LONG_SCALE = re.compile(LONG_SCALE1 + '(.*?)' + LONG_SCALE2, re.S)
        result_LONG_SCALE = pat_LONG_SCALE.findall(buff)
        LONG_SCALE = result_LONG_SCALE[0]

        pat_HEIGHT_SCALE = re.compile(HEIGHT_SCALE1 + '(.*?)' + HEIGHT_SCALE2, re.S)
        result_HEIGHT_SCALE = pat_HEIGHT_SCALE.findall(buff)
        HEIGHT_SCALE = result_HEIGHT_SCALE[0]

        pat_LINE_NUM_COEFF = re.compile(LINE_NUM_COEFF1 + '(.*?)' + LINE_NUM_COEFF2, re.S)
        result_LINE_NUM_COEFF = pat_LINE_NUM_COEFF.findall(buff)
        LINE_NUM_COEFF = result_LINE_NUM_COEFF[0]
        LINE_NUM_COEFF3 = LINE_NUM_COEFF[0]
        LINE_NUM_COEFF3 = LINE_NUM_COEFF3.strip('()')
        LINE_NUM_COEFF3 = LINE_NUM_COEFF3.replace('\n', '')
        LINE_NUM_COEFF3 = LINE_NUM_COEFF3.replace('\t', '')
        LINE_NUM_COEFF3 = LINE_NUM_COEFF3.replace(',', ' ')

        pat_LINE_DEN_COEFF = re.compile(LINE_DEN_COEFF1 + '(.*?)' + LINE_DEN_COEFF2, re.S)
        result_LINE_DEN_COEFF = pat_LINE_DEN_COEFF.findall(buff)
        LINE_DEN_COEFF = result_LINE_DEN_COEFF[0]
        LINE_DEN_COEFF3 = LINE_DEN_COEFF[0]
        LINE_DEN_COEFF3 = LINE_DEN_COEFF3.strip('()')
        LINE_DEN_COEFF3 = LINE_DEN_COEFF3.replace('\n', '')
        LINE_DEN_COEFF3 = LINE_DEN_COEFF3.replace('\t', '')
        LINE_DEN_COEFF3 = LINE_DEN_COEFF3.replace(',', ' ')

        pat_SAMP_NUM_COEFF = re.compile(SAMP_NUM_COEFF1 + '(.*?)' + SAMP_NUM_COEFF2, re.S)
        result_SAMP_NUM_COEFF = pat_SAMP_NUM_COEFF.findall(buff)
        SAMP_NUM_COEFF = result_SAMP_NUM_COEFF[0]
        SAMP_NUM_COEFF3 = SAMP_NUM_COEFF[0]
        SAMP_NUM_COEFF3 = SAMP_NUM_COEFF3.strip('()')
        SAMP_NUM_COEFF3 = SAMP_NUM_COEFF3.replace('\n', '')
        SAMP_NUM_COEFF3 = SAMP_NUM_COEFF3.replace('\t', '')
        SAMP_NUM_COEFF3 = SAMP_NUM_COEFF3.replace(',', ' ')

        pat_SAMP_DEN_COEFF = re.compile(SAMP_DEN_COEFF1 + '(.*?)' + SAMP_DEN_COEFF2, re.S)
        result_SAMP_DEN_COEFF = pat_SAMP_DEN_COEFF.findall(buff)
        SAMP_DEN_COEFF = result_SAMP_DEN_COEFF[0]
        SAMP_DEN_COEFF3 = SAMP_DEN_COEFF[0]
        SAMP_DEN_COEFF3 = SAMP_DEN_COEFF3.strip('()')
        SAMP_DEN_COEFF3 = SAMP_DEN_COEFF3.replace('\n', '')
        SAMP_DEN_COEFF3 = SAMP_DEN_COEFF3.replace('\t', '')
        SAMP_DEN_COEFF3 = SAMP_DEN_COEFF3.replace(',', ' ')
    rpc = ['ERR_BIAS='+ERR_BIAS, 'ERR_RAND='+ERR_RAND, 'LINE_OFF='+LINE_OFF, 'SAMP_OFF='+SAMP_OFF, 'LAT_OFF='+LAT_OFF, 'LONG_OFF='+LONG_OFF, 'HEIGHT_OFF='+HEIGHT_OFF, 'LINE_SCALE='+LINE_SCALE, 'SAMP_SCALE='+SAMP_SCALE, 'LAT_SCALE='+LAT_SCALE, 'LONG_SCALE='+LONG_SCALE, 'HEIGHT_SCALE='+HEIGHT_SCALE,'LINE_NUM_COEFF='+LINE_NUM_COEFF3,'LINE_DEN_COEFF='+LINE_DEN_COEFF3,'SAMP_NUM_COEFF='+SAMP_NUM_COEFF3,'SAMP_DEN_COEFF='+SAMP_DEN_COEFF3]
    #rpc = ['ERR BIAS=' + ERR_BIAS, 'ERR RAND=' + ERR_RAND, 'LINE OFF=' + LINE_OFF, 'SAMP OFF=' + SAMP_OFF,'LAT OFF=' + LAT_OFF, 'LONG OFF=' + LONG_OFF, 'HEIGHT OFF=' + HEIGHT_OFF, 'LINE SCALE=' + LINE_SCALE,'SAMP SCALE=' + SAMP_SCALE, 'LAT SCALE=' + LAT_SCALE, 'LONG SCALE=' + LONG_SCALE,'HEIGHT SCALE=' + HEIGHT_SCALE, 'LINE NUM COEFF=' + LINE_NUM_COEFF3, 'LINE DEN COEFF=' + LINE_DEN_COEFF3,'SAMP NUM COEFF=' + SAMP_NUM_COEFF3, 'SAMP DEN COEFF=' + SAMP_DEN_COEFF3]
    return rpc


if __name__ == "__main__":

    ## 参数设置
    demfile = r'C:\ENVI53\data\GMTED2010.jp2'  # 使用ENVI的DEM文件
    filepath = 'E:\data\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856'  # 解压后的文件夹
    rpbfile = filepath+'\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856.rpc'  #对应的RPC文件
    infile = filepath+'\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856.tiff'
    outfile= os.path.splitext(infile)[0]+'_out.tiff'




    os.chdir (filepath)
    # strip函数是剔除文本中的'\n',split函数是分割规则
    rpc = []
    rpc=read_rpb(rpbfile)


    # dataset = rasterio.open(file)

    dataset = gdal.Open(infile)
    dataset.SetMetadata(rpc, 'RPC')

    dst_ds = gdal.Warp('outfile.tif', dataset, dstSRS='EPSG:4326', xRes=0.0002, yRes=0.0002,
                       rpc=True, transformerOptions=[r'RPC_DEM=' + demfile])

---

测试1

原始数据：

rpc校正后数据:

---

小结

雷达影像的相机坐标与光学影像相机坐标是不一样的。

光学卫星相机坐标是符合正常的拍摄逻辑，但是雷达卫星的相机坐标是上下颠倒、左右颠倒。通过无GCP点的PRC校正，可以初步将地理坐标赋予给雷达影像。

代码中的参数需要根据不同电脑重新设置，其他无需改动即可进行RPC校正。

## 参数设置
demfile = r'C:\ENVI53\data\GMTED2010.jp2'  # 使用ENVI的DEM文件
filepath = 'E:\data\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856'  # 解压后的文件夹
rpbfile = filepath+'\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856.rpc'  #对应的RPC文件
infile = filepath+'\GF3_MYN_SL_021996_E113.7_N22.0_20201013_L1A_HH_L10005135856.tiff'
outfile= os.path.splitext(infile)[0]+'_out.tiff'