1

Topic: Subset Pléiades tie point selection

Dear all,

I created a DEM (1m) from subsets of Pléiades tristereo images (acquisition date 9/07/2013). It was necessary to create the DEM based on the subsets in order to attain a y parallax of 2 (When using the entire Pléiades image, I could not even reach a y parallax under 7). Now I would like to use this DEM to compare 2 Pléiades images (1: acquisition date 9/07/2013, 2:  acquisition date 16/07/2012), but the DEM is smaller (a quarter) than both Pléiades images. Also, these two Pléiades images only overlap for half of their area (in this area, the DEM is created). It is not possible to create an ancillary file of subsets of the two Pléiades images conform to the DEM subset area in COSI-Corr.

What I could do is create ancillary files for the two entire Pléiades images but my question is now if it makes sence to select tie points in the Pleiades image and the subset DEM, so only having tie points in the pleiades image for the subset area? Or is there another way in which I would be able to analyse ground deformations in only a part of a Pléiades image comparing it to a DEM that is smaller than the Pléiades image ? It would be problematic for my further analyses if I should always work with the entire Pléiades images, as I will calculate the DEMs with Pléiades images for which I need to take subsets to reach a good y-parallax.

Exelis technical support ensured me that "You can definitely use the subset images that contain RPC information and xstart and ystart values in any ENVI tools that require RPC information (orthorectification, DEM extraction, etc.). If the pixel size was not adjusted during the subsetting, then the RPC information does not need to be changed from the RPC information in the original image. You will notice in the header of these subset images that there is an xstart and ystart value. These values tell ENVI where the subset is in relation to the original image. This information helps ENVI understand how to calculate correct locations for each pixel in the subset from the RPC information of the original larger image."

Thank you for your support.
Best regards,

Elise

2

Re: Subset Pléiades tie point selection

Dear Elise,

your situation is common and not problematic.

Generate the ancillary on the full raw image, always. Never, ever, crop a raw image - you'll do the cropping during the orthorectification when you define the ground output grid.

In your situation you'll basically follow these steps:
- generate the ancillary files on the raw images
- For the first pleiade image, select tie-points between whatever reference you have ONLY on the area where you have DEM coverage - Spread them as much as you can as long as you stay withing DEM coverage.
- Optimize the tie-points/GCPS  and orthorectify the first image (you can choose what part of the image you want to orthorectify during that step - this is where you "crop" the image)
- For the second pleiade image, select tie-points ONLY on the common subset with the first pleiade image ortho AND ONLY on areas where you have DEM coverage
- Optimize tie-points/GCPS and orthorectify (usually on the same footprint than the first pleiade ortho)

Note that when you define the output ground footprint during the orthorectification, it is not really a problem is some of the footprint is outside your area of tie-points selection or outside the DEM coverage, you'll just have to keep in mind that this area in the orthoimage won't be trustworthy.

Best,
Francois

3

Re: Subset Pléiades tie point selection

Thank you for the detailed description François !