Gpt: geographic conversion

snap
#1

Dear all,

I would like to understand how gpt converts latitude/longitude values to x/y points.

Thanks in advance for your collaboration,
Monica

#2

Do you mean the row and colum number of a pixel?
What data are you talking about?

#3

Exactly, the x and y number of a pixel used on pin locator and gpt (to do pixel extractions automatically).
SEN3 Olci.

#4

That means you set a pin on a pixel and want to know which position (x/y) in the raster it has?
Or do you want to extract the pixel value at a given lat/lon coordinate for multiple images?

#5

I would like to understand how the system translate my input lat/lon to pixel x/y.

#6

I’m not sure but I think the opposite is the case. Each pixel with x and y can be assigned to a geographic coordinate in latitude and longitude. You can use the tie-point grids in the product to read the corresponding lat and lon value of any pixel.

grafik.png1919×593 798 KB

#7

Thanks for your answer…but the question is: how gpt determine where is exactly degrees coordinates? Using pixel info you can see exactly the center of the pixel but when I asked a pixel towards the edge, which pixel gpt return me? To retrieve this in my code, I used the haversine formala, that return the great-circle distance between two points on a sphere but I would like to understand how gpt determine these values x and y.

#8

Hi Monica,

first, you should use SNAP and not gpt as name for the application. gpt is just one command line tool provided by SNAP.

However, there are different ways from lat/lon coordinates to x/y coordinates.
It depends on which data you are using. Sometimes there is with each pixel x/y a lat/lon coordinate given, sometimes there is only a sparse tie-point grid and in between the tie-points we need to interpolate the lat/lon coordinates. And then are cases were an affine transform is available to compute the lat/lon for a given x/y.

After retrieving the x/y for a given lat/lon, the values at x/y is returned, even if the location is at the edge of the pixel.
We don’t use the great-circle distance between two points here.

S3A and GPT
#9

Thanks for your answer :wink:
Using the word gpt, I’m referring to gpt with pixex module.
In any case, are you able to give me the method implemented using S3A raster data sets?

#10

you can find the code for the pixel-based approach here:


and the tie-point-based here: