tl;dr guide

This page is designed to provide a “bare bones” user guide to spymicmac/MicMac for processing KH-9 Hexagon Mapping Camera images or scanned air photos. It is presented primarily as a list of commands to run; if you would like a longer explanation of the different steps, see the additional guides here.

KH-9 processing

To begin, run generate_micmac_measures to generate id_fiducial.txt and MeasuresCamera.xml.

Next, create a new directory, Ori-InterneScan, and move MeasuresCamera.xml into it.

Finally, run spymicmac.micmac.create_localchantier_xml() to create MicMac-LocalChantierDescripteur.xml, or download the example provided.

Note

The rest of this guide assumes that you have the following directory structure:

project
├── id_fiducial.txt
├── Img1_a.tif
├── Img1_b.tif
├── Img2_a.tif
├── Img2_b.tif
...
├── MicMac-LocalChantierDescripteur.xml
├── Ori-InterneScan
│   └── MeasuresCamera.xml

where Img1_a.tif and Img1_b.tif are the two halves of one of the KH-9 images, Img2_a.tif and Img2_b.tif are two halves of the next image in the acquisition, and so on. For stereo processing, you need a minimum of two images.

joining image halves

To join image halves use the command line tool join_hexagon:

join_hexagon

This will search for all files that fit the pattern DZB*N001_(a|b).tif, and join them into a single image, DZB*N001.tif. After joining the images, create a new directory, halves, and move the image halves into it.

reseau grid

Next, use find_reseau_grid to find the location of the 1081 Reseau marks in the image:

find_reseau_grid DZB*.tif

This will create a file, Ori-InterneScan/MeasuresIm-DZB*N001.tif.xml, for each of the images in the directory. It will also create an image in match_imgs that shows the estimated distortion pattern.

removing crosses (optional)

To erase the Reseau marks from each image, run remove_crosses:

remove_crosses DZB*.tif

This will move the original images to a new directory, original, and save the erased images to the working directory.

resampling

After this step, you can use resample_hexagon to resample the images to the same size, and correct the distortion using the Reseau marks:

resample_hexagon DZB*.tif -s SCALE

where scale is the scale of the resampled image, in pixels per mm (default is 70). This will create a new file, OIS-Reech_DZB*N001.tif for each of the original images.

general micmac processing

Once the images have been re-sampled, the rest of the workflow is largely the same for both KH-9 images and scanned air photos.

tapioca

Once the images are re-sampled, run mm3d Tapioca MulScale to find tie points in downscaled versions of the images:

mm3d Tapioca MulScale "OIS.*tif" 400 1200

In the above command, the numbers after “OIS.*tif” are the size, in pixels, of the longest dimension of the downscaled image.

tapas

To find the relative orientation of the images, and calibrate the camera parameters, use mm3d Tapas:

mm3d Tapas RadialBasic "OIS.*tif" Out=Relative LibFoc=0

The LibFoc=0 flag will keep the focal length fixed at the value provided in MicMac-LocalChantierDescripteur.xml.

relative dem and orthomosaic

Create a relative orthophoto and DEM using mm3d Malt:

mm3d Malt Ortho "OIS.*tif" Relative DirMEC=MEC-Relative NbVI=2 ZoomF=8 DefCor=0 CostTrans=1 EZA=1

If you have used an image mask, run the following command instead:

mm3d Malt Ortho "OIS.*tif" Relative DirMEC=MEC-Relative NbVI=2 MasqImGlob=filtre.tif ZoomF=8 DefCor=0 CostTrans=1 EZA=1

By default, mm3d Malt only orthorectifies the individual images; to create an orthomosaic, use the mm3d Tawny command:

mm3d Tawny Ortho-MEC-Malt Out=Orthophotomosaic.tif RadiomEgal=0

Note

If the image is very large, you may need to run mosaic_micmac_tiles.py to combine the tiles into a single image. For the relative DEM, this will probably not be needed.

For the orthophoto, run the following from the Ortho-MEC-Relative directory:

mosaic_micmac_tiles.py -filename Orthophoto

registering the image

Note

This step requires a DEM and an orthoimage to find control points and estimate the absolute orientation of the images. In the examples below, I assume that these are named DEM.tif and Landsat.tif, respectively.

The main command to run here is register_relative:

register_relative MEC-Malt DEM.tif

Note

If you have a shapefile of the image footprints, use the -footprints flag; otherwise, they will be downloaded from USGS Earth Explorer:

register_relative MEC-Malt DEM.tif -footprints Footprints.shp

The shapefile should have a polygon for each image, with the name of the original image (minus the file extension) included in an ID column.

dem and orthomosaic

After the absolute orientation has been estimated by registering the image, run mm3d Malt again with this new orientation to extract the final DEM and orthophoto:

mm3d Malt Ortho "OIS.*tif" TerrainFirstPass DirMEC=MEC-Malt NbVI=2 ZoomF=1 DefCor=0 CostTrans=1 EZA=1

If you have used an image mask, run the following command instead:

mm3d Malt Ortho "OIS.*tif" TerrainFirstPass DirMEC=MEC-Malt NbVI=2 MasqImGlob=filtre.tif ZoomF=1 DefCor=0 CostTrans=1 EZA=1

To generate an orthomosaic, run the following command:

mm3d Tawny Ortho-MEC-Malt Out=Orthophotomosaic.tif RadiomEgal=0

Note

If the image is very large, you may need to run mosaic_micmac_tiles.py to combine the tiles into a single image. For the DEM, run the following from the MEC-Relative directory:

mosaic_micmac_tiles.py

And for the orthophoto, run the following from the Ortho-MEC-Relative directory:

mosaic_micmac_tiles.py -filename Orthophoto

You can also run post_process_micmac to apply the AutoMask to the DEM and Orthomosaic, and georeference the correlation mask.

And that’s it. You should now have an orthorectified KH-9 (or air photo) mosaic, and a DEM. Enjoy.