topic Re: Converting raster XY co-ordinates to lat/long in ArcGIS Spatial Analyst Questions

Converting raster XY co-ordinates to lat/long

LamaZip — Sat, 11 Dec 2021 15:29:39 GMT

I have downloaded a SST image of the United Kingdom. It is in PNG format with no coordinate system.

I know the dimensions of my image and its lat/long range. However, i'm not sure what approach to take in Arcmap to set the co-ordinate system.

There is a web-page describing how to convert image co-ordinates to latitude/longitude but it is beyond my scope link

The image is apparently produced in Mercator projection

Note, this will need to be carried out on multiple rasters, so a automated/batch approach, which isn't interactive or manual, is required

1. is there anyway to automatically convert the XY pixel co-ordinates to lat and long, by knowing the lat/long range of the image and the dimensions?
2. If not, how would one do this calculation to convert image XY co-ordinates to lat/long in Arcmap / Python?

Details of Image

Dimension   1596 x 1617

BPP 8

Latitude range  47 — 62.999108
Longditude range    -15 — 13
Pixel dimensions    1.100149 x 1.100149

Re: Converting raster XY co-ordinates to lat/long

ChrisDonohue__GISP — Mon, 29 Aug 2016 18:18:08 GMT

I saw this in the documentation, which may resolve part of the puzzle. As for the rest, though....not sure.

What datum and ellipsoid do you use to map the data?

Panorama uses a sphere of diameter 6370718 meters to approximate the Earth's surface. This is used as the datum baseline for all mapped processing. The regional Molodensky offsets between the sphere and a WGS 84 ellipsoid based datum are available on request.

https://www.neodaas.ac.uk/faq/#datum

Adding Melita Kennedy

Chris Donohue, GISP

Re: Converting raster XY co-ordinates to lat/long

DarrenWiens2 — Mon, 29 Aug 2016 19:06:00 GMT

You can place the raster semi-manually by creating a world file. For your example file (janW1.png), the world file name would be janW1.pngw.

Here is an example using the following world file (lines 5 & 6 should be adjusted by half a pixel to move to the pixel center):

0.017543859
0.0
0.0
-0.009894315
-15.0
62.999108

...and here is the result, although it doesn't match exactly. This may be due to cdspatial‌'s note about a funky ellipsoid (this is over WGS84):

Re: Converting raster XY co-ordinates to lat/long

MelitaKennedy — Mon, 29 Aug 2016 20:20:00 GMT

Even though they say it's in Mercator, there's a bunch of north-south distortion at those latitudes. The Y range is almost 3 times the X range in Mercator (approximately--I didn't use the sphere values) so as Darren demonstrated, it looks like the raster is really in lat/lon (geographic CRS).

I'm also attaching a Panorama prj file. You would script in a call to Define_Projection to set this coordinate system.

You may need to also define a transformation to pick up the difference between the sphere and WGS84. You can use the Create Custom Geographic Transformation tool. Use the geocentric translation method and leave the parameter values set to zeroes.

You might also query the data provider for the "regional molodenky transformation values." If the sphere model is earth-centered, the Molodensky parameters should be zeroes unless they're making some other kind of adjustment.

Melita

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Tue, 30 Aug 2016 06:37:40 GMT

Hi Melita.

I'm not sure if i'm doing this correctly but I selected my image, then went to raster -> define projection ->. Selected my PNG as the input and imported GCS_panorama as the coordinate system.

However, this is the resulting image. Not sure what's happened but it's very different from the original.

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Tue, 30 Aug 2016 06:43:22 GMT

Hi Chris.

I manually converted the XY coordinates to lat and long using the calculation provided on the webpage.

I assumed I had done the calculation wrong, but luckily this world file shows that it is correct as we have got the same result.

Could you explain how to `(lines 5 & 6 should be adjusted by half a pixel to move to the pixel center):`

and why it would be necessary?

Also do you have any tips for the aligning the offset? Thanks.

Re: Converting raster XY co-ordinates to lat/long

ChrisDonohue__GISP — Tue, 30 Aug 2016 16:24:47 GMT

Punting this back to one of the experts: Darren Wiens

Chris Donohue, GISP

Re: Converting raster XY co-ordinates to lat/long

DarrenWiens2 — Tue, 30 Aug 2016 17:07:54 GMT

(lines 5 & 6 should be adjusted by half a pixel to move to the pixel center)

Lines 5 and 6 in the world file are supposed to reference the coordinates of the upper left pixel center, but I was lazy and just used the upper left coordinates of the raster (so, the coordinates correspond to the upper left corner of the upper left pixel). Because the raster is shifted by much more than a pixel, being off by half a pixel doesn't really matter, for now.

Re: Converting raster XY co-ordinates to lat/long

MelitaKennedy — Tue, 30 Aug 2016 22:24:29 GMT

What's the data frame's coordinate system? And the raster's extent after the define projection?

Thanks!

Melita

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Wed, 31 Aug 2016 18:11:56 GMT

Hi Melita,

Extent of original (No spatial reference)

Extent after GCS_panorama has been applied using `define projection`. For some reason the whole image becomes skewed, or flipped, or something.

Running the mouse over the image...
bottom left corner is -186, -85 Decimal degrees

bottom right corner is 177, 86
top left corner is -179, 0 degree

top right corner is 180, 0 degree

This error appears if I removed the PNG and re-add after GCS_panorama has been applied

Any thoughts?

Thanks,

Prem

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Wed, 31 Aug 2016 18:13:07 GMT

I see. Thanks for the explanation!

Any idea on what I can do to shift the PNG into the correct co-ordinates? It's so close, yet so far.

Re: Converting raster XY co-ordinates to lat/long

MelitaKennedy — Wed, 31 Aug 2016 22:16:23 GMT

You've got to do both--make sure the world file is applied to convert the raw pixel coordinates to latitude/longitude values and define the Panorama geographic coordinate system so that the system understands the earth model that's being used. I think once you do that, AND set a custom geographic transformation (between Panorama and WGS84, method: geocentric translation, parameters: zeroes), it will line up.

Defining Panorama without using the world file means that the raster's extent values wrap around the entire world (they're greater than 360 degrees) thus the weird display.

See this topic about world files for more information.

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Thu, 01 Sep 2016 09:40:23 GMT

Just attempted and it didn't seem to work. I may of chose the wrong WGS84 coordinate system? Here's my work flow.

1. Add janw1.png with world file

2. Define geographic coordinate system

Define Projection tool -> janw1.png -> Panorama

3. Create custom geographic transformation

input = panorama

output = wgs_1984 (is this the correct WGS84?)

method = geocentric. All values zero.

4. Project Raster

input = janw1..

output coordinate sys = wgs_1984

resampling = nearest

5. The output has marginally moved upwards, but still doesn't align.

I have a feeling I might of missed a step, or chosen the wrong coordinate system?

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Thu, 01 Sep 2016 09:54:52 GMT

Just attempted with geographic transformations between Panorama and...

GCS -> WGS_1984
Projected ->WGS_1984_Web_Mercator_Auxiliary_Sphere

Projected -> WGS_1984_World_Mercator

The results all seem to be identical.

Re: Converting raster XY co-ordinates to lat/long

LamaZip — Wed, 07 Sep 2016 17:41:49 GMT

Hi Melita.

Any idea on what I could be doing wrong? I've still been tinkering away without much luck.

Any advice would be greatly appreciated!

Thanks