8.22. SPOT5

SPOT5 is a CNES (Space Agency of France) satellite launched on May 2002 and decommissioned in March 2015. SPOT5 contained two High Resolution Stereoscopic (HRS) instruments with a ground resolution of 5 meters. These two cameras were pointed forwards and backwards, allowing capture of a stereo image pair in a single pass of the satellite.

ASP supports only images from the HRS sensors on SPOT5. These images come in two parts, the data file (extension .bil or .tif) and the header file the data file (extension .dim). The data file can be either a plain binary file with no header information or a GeoTIFF file. The header file is a plain text XML file. When using SPOT5 images with ASP tools, pass in the data file as the image file and the header file as the camera model file.

All ASP tools can handle .bil images (and also .bip and .bsq) as long as a similarly named .dim file exists that can be looked up. The lookup succeeds if, for example, the .dim and .bil files differ only by extension (lower or upper case), or, as below, when an IMAGERY.BIL file has a corresponding METADATA file.

A sample SPOT5 image can be found at at http://www.geo-airbusds.com/en/23-sample-imagery.

One issue to watch out for is that SPOT5 data typically comes in a standard directory structure where the image and header files always have the same name. The header (camera model) files cannot be passed into the bundle_adjust tool with the same file name even if they are in different folders. A simple workaround is to create symbolic links to the original header files with different names:

bundle_adjust -t spot5                                            \
  front/SEGMT01/IMAGERY.BIL back/SEGMT01/IMAGERY.BIL              \
  -o ba_run/out
parallel_stereo -t spot5                                          \
  front/SEGMT01/IMAGERY.BIL back/SEGMT01/IMAGERY.BIL              \
  st_run/out --bundle-adjust-prefix ba_run/out

See Section 6 for a discussion about various speed-vs-quality choices of the stereo algorithms.

One can also mapproject the SPOT5 images before they are passed to parallel_stereo. In order to do so, you must first use the add_spot_rpc tool to generate an RPC model approximation of the SPOT5 sensor model.

add_spot_rpc front/SEGMT01/METADATA.DIM -o front/SEGMT01/METADATA.DIM
add_spot_rpc back/SEGMT01/METADATA.DIM  -o back/SEGMT01/METADATA.DIM

This will append the RPC model to the existing file. If the output is a separate file, only the RPC model will be saved to the new file.

Then use the spot5maprpc session type when running parallel_stereo on the mapprojected images. See the note in Section 6.1.7 about perhaps reducing the resolution of the DEM to mapproject onto if ghosting artifacts are seen in the produced DEM.

mapproject --tr gridSize sample_dem.tif front/SEGMT01/IMAGERY.BIL   \
  front/SEGMT01/METADATA.DIM front_map_proj.tif -t rpc
mapproject --tr gridSize sample_dem.tif back/SEGMT01/IMAGERY.BIL    \
  back/SEGMT01/METADATA.DIM back_map_proj.tif -t rpc
parallel_stereo -t spot5maprpc front_map_proj.tif back_map_proj.tif \
  front/SEGMT01/METADATA.DIM back/SEGMT01/METADATA.DIM              \
  st_run/out sample_dem.tif

Notice how we used the same resolution (option --tr) for both images when mapprojecting. That helps making the resulting images more similar and reduces the processing time (Section

See Section 6 for a discussion about various speed-vs-quality choices of the stereo algorithms.


Fig. 8.21 Cropped region of SPOT5 image and a portion of the associated stereo DEM overlaid on a low resolution Bedmap2 DEM.