I have heaps of aerial photos that need to be georeferenced.

I have access to ArcMap and ERDAS.

All I am given is an excel sheet with the centroid coordinates of each image.

Each photo covers an 18m x 24m region.

The photos are in a very remote region with no roads or other structures to georeference them to.

Is it possible to do this?


If you know the image size (pixels) and scale you can work out the top left corner from the centroid. You can use Excel to do the math. Then create a txt list and create world files for each image.

I would do one manually in ArcGIS.

To georeference one see: http://library.columbia.edu/indiv/dssc/eds/georef.html

Then you can use the values in Excel to pixel scale.

32.0 (pixel size)
0.0 (rotation)
0.0 (-rotation)
-32.0 (-pixel size)
691200.0 (x coordinate top left)
4576000.0 (y coordinate top left)
| improve this answer | |
  • 1
    Rotation is the really difficult part here, or crabbing rather. Sometimes the crab is not known, and it is not that easy to measure precisely off the image. This gets even worse if these are collared images with a horizontal pixel gridding imposed on crabbed aerial photos. – blord-castillo Jul 6 '11 at 17:33

To start you could use one of the ESRI map service (world imagery) basemap layers to georeference off of. Depending on the date of your aerials you will have to use natural features to georeference to (e.g. forest edges or stream channels).

| improve this answer | |
  • Yeah but it uses Web Mercator and then you would have to reproject it to a different projection . – PROBERT Dec 8 '17 at 1:39
  • @PROBERT, why would you have to reproject it to a different projection? – artwork21 Dec 8 '17 at 12:42
  • some people may not know this so that is why I was just saying...so just to make sure they change to a different projection so their work do not messed it up. – PROBERT Dec 8 '17 at 16:27
  • Users could also change the data frame CRS to whatever and then georeference bypass having to re-project raster afterwords. – artwork21 Dec 8 '17 at 16:29
  • For the basemap, I think it will be rough to do that... – PROBERT Dec 8 '17 at 16:45

Since each individual photo has such a small footprint, you could use the open source photo-stitching software Hugin to stitch a few tiles together. I have used Hugin (not for aerial photos, but for landscape photography), and it's incredibly good at tiling images. Then you'd have multiple points per scene to georeference on.

| improve this answer | |
  • 2
    That seems like a good idea, but you'd have to use an image editor to mark the centre of each photo first, otherwise you wouldn't know which point to georeference. – Ian May 10 '12 at 4:21
  • @Ian - nice idea..the control points from marked stitched photo can be reused for the unmarked image.. – vinayan May 10 '12 at 7:18

Do you have an index that illustrates how the images lie in relation to one another? If so, you may be able to mosaic them into one image based on that index, then georeference the resulting image using the centroids of each original image.

One caveat: if adjacent images are supposed to overlap each other then this approach won't work. You'll end up with repeated features and a mosaic that is larger than the area the input images represent.

| improve this answer | |

I believe you would need to have the scale and the route of the aerial photos, and also the fiducial points (which are the corners of the photo). And yet at least 2 or 3 more reference points would have to be used. I think the idea of having the centroid coordinates is only to help you with a starting point to the process of georreferencing.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.