Skip to content

Table of content

Tips and Tutorials \ Generate an ortho-image (PNEO)

Generate an ortho-image (PNEO)

This tutorials explains how to generate an ortho-image with the example of PNEO data

Tools:
Difficulty:
Type: Tip

Overview

This tutorial shows how to generate ortho-images from PNEO data but it can be applied to other sensors.

Pre-requisites

This tutorial uses command lines from the OTB.

Commands

Command list :

1 - Get incidence angles (this command is to be launched in every folder, because we will work in the folder where the value in minimal) 

grep INCIDENCE_ANGLE *DIM*PAN*

2 - Pansharpening

otbcli_BundleToPerfectSensor -inp <path>/<pan.[tif|jp2]> -inxs <path>/<[rgb|ned].[tif|jp2]> -out <pxs.tif>

2bis (opt) - Refine sensor model

otbcli_HomologousPointsExtraction -in1 <pxs>.tif -in2 <ortho_ref>.tif -mode geobins -elev.dem /work/datalake/static_aux/MNT/SRTM_30_hgt -algorithm sift -out homologous.txt -outvector homologous.shp

otbcli_RefineSensorModel -ingeom <input>.geom -outgeom <output>.geom -elev.dem /work/datalake/static_aux/MNT/SRTM_30_hgt -inpoints homologous.txt -outstat <stats>.txt -outvector <refined>.shp -map epsg -map.epsg.code 32631

3 - Ortho-rectification 

otbcli_OrthoRectification -io.in <path>/<pan+XS.tif> -io.out <ortho.tif> uint16 [-elev.dem <MNT.tif>]

Details

Image selection

In case you are working with stereo or tri-stereo acquisition, you have more than 1 image for the same scene. In order to select which acquisition is to be used to generate the ortho-image, we look at the incidence angle for each image (here the panchromatic one), then we will work with the acquisition where the incidence angle is minimal (i.e the closest to a nadir shot).

In every folder, run the following command to read the incidence angle in the DIMAP file :

grep INCIDENCE_ANGLE\ u *DIM*PAN*

Pansharpening

Pleiades-Neo provides XS images in 2 different files :

  • IMG_PNEO*MS-FS*RGB*.JP2 : the first containing RGB bands
  • IMG_PNEO*MS-FS*NED*.JP2 : the second containing near infrared, red-edge and deep-blue bands

We can therefore apply pansharpening method from the OTB (XS + PAN fusion) for each image :

ml otb
otbcli_BundleToPerfectSensor -inp <path>/<pan.[tif|jp2]> -inxs <path>/<[rgb|ned].[tif|jp2]> -out <pxs.tif>

Refine sensor model (optionnal)

In order to improve the localization of the ortho-image, we can refine the sensor model. To do so, we have to compute homologous points between P+XS and a reference ortho-image :

otbcli_HomologousPointsExtraction -in1 <pxs>.tif -in2 <ortho_ref>.tif -mode geobins -elev.dem /work/datalake/static_aux/MNT/SRTM_30_hgt -algorithm sift -out homologous.txt -outvector homologous.shp

Then the model can be refined :

otbcli_RefineSensorModel -ingeom <input>.geom -outgeom <output>.geom -elev.dem /work/datalake/static_aux/MNT/SRTM_30_hgt -inpoints homologous.txt -outstat <stats>.txt -outvector <refined>.shp -map epsg -map.epsg.code 32631

If you do not have a <input>.geom file, you can use the following command to generate it :

otbcli_ReadImageInfo -in <DIMAP_PAN>.xml -outkwl <input>.geom

Ortho-rectification

In case you have a DEM for the targeted area you can execute the following command :

otbcli_OrthoRectification -io.in <path>/<pan+XS.tif> -io.out <ortho.tif> uint16 -elev.dem /work/datalake/static_aux/MNT/SRTM_30_hgt
Else, the OTB will use the geoid as reference :

otbcli_OrthoRectification -io.in <path>/<pan+XS.tif> -io.out <ortho.tif> uint16

Tips inside the tip !

  • If you have a DEM with a higher resolution than SRTM (DEM Pleiades, DEM Pleiades-Neo, etc.) for the observed area, it is better to use it during orthorectification.

  • In order to perform the refinement of the sensor model, reference ortho-images can be found on the IGN website, which provides open-access to the BD Ortho.

Troubleshooting

  • During the refinement of the sensor model, in the homologous points extraction step, make sure the compared bands are the same (Red with Red for example). Otherwise, it is necessary to map bands with the OTB commandline (by default, the command compares band 1 of each image).

  • You can sometimes observe a localization error with P-Neo images during orthorectification. You can use an external VT (OSM, etc.) to check if this bias appears or not.

References

Contacts

    Questions ? Get help on the Forum