Rectification for Rapideye imagery
We are going to have a look at GeoRectification of data from the high resolution sensor RapidEye. We are lucky with this lab in that we can rely on improvements in sensor technologies to rectify the data at hand. Do you recall the methods we employed in Geog 300 to rectify an aerial photo? Perhaps we should practice this technique using PCI?!
Roger has created a PCI image from the Rapid Eye data we received for Prince George. You can see the original downloads as well as the pix file in the /home/labs/geog457/lab4-rectification/rapid_eye_pg_area directory.
Points to think about:
- What is the file format the files come in
- What was the process that the files were loaded into the pix file – we should all be pros at this by now
- What are the other non raster layers (segments in the file).
Look at the original data
Load the pix file in the /home/labs/geog457/lab4-rectification/rapid_eye_pg_area directory using the file and north to the top options. Now load the pg boundary shape file and the DEM raster layer in the /home/labs/geog457/lab4-rectification directory. How do things line up?
Now load the subset of the Rapid Eye data Scott created from the DEM extents into the project (pg2013re_city_bound file). How does this layer look? What is going on here?
As well as the in being much easier than what we did in Geog300, we are going to include a DEM to OrthoRectify our imagery (to compensate for changes in relief). You need to open up OrthoEngine and move through the steps of creating a project, loading data and completing the orthocorrection by:
- Create a project (new project in your lab4-rectification directory)
- Pick the proper math model – you have to check out the options under the radio buttons
- Set the projection to UTM Zone 10 NAD 83 (pick the most appropriate projection from the list)
- Data input – we only have one uncorrected image to load – load the pg2013re_city_bound.pix file
- We will skip all the other steps (as we are not using any GCPs or any other layers)
- Select Ortho generation –> schedule a generation –> =make sure you pick a DEM (select the channel when it loads)
Your panel should look something like the image below (click it):
Then run the ortho rectification and look at the results in Focus.
Why is everything so bizarre looking? Lets discuss the steps that led to the output.
Now Try it again but use the eye_ball pix file Scott created. Lets discuss what happened this time.
Look Through the Data you Created and run through some of the tasks Roger was thinking we should look – see the following notes:
- 321 composite (natural colour)
- 532 (infra-red)
- 543 (NIR -Red Edge – Red)
Check the DNs in the display
- most features seem to have DNs that either all increase or all decrease through the 3 bands
- e.g. water – low DN decreasing 3-4-5
- forest: high DNs increasing 3-4-5 Red edge alone in grayscale
- which features have the highest values ? (soccer fields?)
- what is correlation between 5 and 4, tools -> scatterplot
- 4 and 3 (how is red edge correlated with NIR and Red respectively … which is it closer to ?
- View histograms: 1. visible bands are single-peak RE and NIR are bi-modal
- more pronounced in NIR (why)?
- Note minimum value in each band – it progressively gets lower. WHY ?
- It is because of haze – haze increases the DNs in blue, less so in green and so on as wavelength increases Band ratios
Ratios – in raster calculator run these ratios:
- in display only, and to 32 bit display to see the decimal values 5/4
- 5/3 (traditional vegetation ratio)
- 4/3 (RE/Red) they all seem to be trimodal – what are the 3 peaks ? (Roger thinks they are water, unvegetated e.g. urban, and vegetated – remember the 5, 4,3 composite/stretch)