Showing results for 
Search instead for 
Do you mean 

Geometrically Calibrate Pleiades DIMAP with RPCs

by Technical Evangelist on ‎12-21-2018 08:39 AM - edited on ‎02-21-2020 06:49 AM by Community Manager (926 Views)


Download models


Original positional accuracy is shown in the lefthand View, increased spatial accuracy when utilising the RPCs on the right


Some raster format readers in ERDAS IMAGINE 2018, such as the NITF and DigitalGlobe TIL format readers, will automatically identify the presence of RPC information specific to the particular image so that the software recognizes the data as using a 3D RPC-based geometric model (so that the data can be orthorectified on the fly, used as part of a stereo-pair in Stereo Analyst for ERDAS IMAGINE, fine-tuned in IMAGINE AutoSync etc) . However other format readers, such as those for DIMAP v1 and v2 based formats (SPOT 6, SPOT 7, Pleiades, KazEOSAT, etc)  do not automatically recognize the associated RPC information. Instead the user must manually geometrically calibrate (associate) the RPC information to the image so that ERDAS IMAGINE can use that information for accurately geopositioning the data.  This manual association is usually performed one image at a time via the Raster tab > Geometry group > Geometric Calibration menu > Calibrate with Sensor Model and Terrain option.




The first Spatial Model (attach_rpc_calibration_v16_5_0.gmdx) illustrates the general methodology for performing this task in Spatial Modeler. The user must identify the input raster image filename, the associated RPC information file and the type of sensor involved. Optionally an input DEM source can be identified as a terrain source to be used by the RPC model. This field can be left blank if the average terrain height specified by the RPC file is to be used.


Also optionally, the user can name a proxy file to be created, rather than associating the updated geometric calibration with the Input Image (which isn't always possible). The proxy file is simply a pointer to the original data, but with the geometric calibration associated and can be used as if it were an image. If the Proxy field is left blank the RPC geometric calibration will be associated with the original Input Image. A proxy file can be used in all ways like a regular image - it is basically a small header file pointing to all the necessary files from the original data. It can be displayed in a 2D View, input to any raster processing functions in ERDAS IMAGINE, etc. It's advantage is that the original image can still be used in its original state if there turns out to be a problem with what was added via the proxy file.


The Spatial Model also tests to see if the input already has a 3D geometric model associated with it - if it does then it does not attempt to alter the existing information.


You could easily make it more general by scanning input directories for images, building RSETs, etc. To assist in automating this process for multiple images, the second Spatial Model provided (attach_rpc_calibration_pleiades_v16_5_0.gmdx) shows how to geometrically calibrate Airbus Pleiades DIMAP v2 images (e.g. so you can use them as stereo pairs in SAfI). Airbus Pleiades DIMAP v2 images have a naming convention following the pattern DIM_PHR<identifier>.XML, which is a header pointing to the actual image files that are usually stored as TIFF or JPEG 2000 files. There will also be other ancillary files, including a RPC_PHR<identifier>.XML which contains the RPC information.




The user is promoted to provide a directory location - that directory and all subdirectories will be searched for input images matching the DIM_PHR*.XML pattern. Modify the Multi Filename Input operator if you want other patterns.


There are also input options as to whether you want statistics and RSET pyramids to be generated. Statistics are useful for correct display of the data and RSETs assist in rapid zooming of the data. If the input images are pointing to JPEG 2000 data then you don't need to generate RSETs (pyramids are intrinsic to the JPEG 2000 encoding scheme).


Finally there is a port on the Iterator setting the use of an input DEM to be associated with the RPC information. This is actually optional (as per the first Spatial Model above) but has been hardwired in the provided Spatial Model to use the Terrain Library shipped with ERDAS IMAGINE. Alter this to suit the DEM data you have available.


The Multi Filename Input operator will generate a list of valid input images. This list is passed to an Iterator which will apply its processing to each of the input images.


Iterator sub-model


Inside the Iterator, each input DIM_PHR*.XML name is broken down in order to automatically identify the associated RPC file for each image, and to create a name for the output proxy file (which is set to DIM_PHR*_proxy.uai in the same directory as the input image). Alter to suit your needs. 


The Sensor Model Input is hardwired to Pleiades RPC. Again this could be altered to suit the data you have to calibrate.


As each iteration creates a new proxy header a message will be logged in the Session Log as an indicator of progress and to help identify if the Spatial Model finishes early because of a problem.


 Pleiades image pair being exploited in Stereo Analyst for ERDAS IMAGINE after associating RPC models



 Note that this Spatial Model could be easilly combined with the Geodetic Reproject model if the desired result is a new orthorectiified iimage file in the Projected Coordinate System of choice.