Mosaic JPEGs

[VarunShankar]

Hi,

I'm utterly new to ArcGIS. I could use some help with what I need to accomplish. I have a folder full of aerial photographs shot as a plane flew over a field in one direction, then the reverse and so on, several times. I need to:

1. Stitch the JPEG images of these flybys together into a mosaic.
2. Export the assembled mosaic into a JPEG.
3. Orthorectify this assembled mosaic.

I'm having trouble even with step 1. I imagine I could break step 1 down further and stitch a single image for each direction the airplane flew, and then stitch those images together. I have broad GPS data for the field. I simply don't know how to use ArcMap to accomplish what I need!

I'd really appreciate any help or links. This is a super-time-crunch project that I probably shouldn't have left for so late 😞

Thanks!

[RobertBerger]

Greetings,

So the first question is: do the air photos already have a spatial reference or are they raw?
If your photos are raw then the next question is what kind of camera was used and what kind of metadata is available. If you have a professional air photo camera then you can look into using the camera frame raster types as part of mosaic dataset loading.
If all you have are some pictures someone took with their camera out of a plane, then you'd have to georeference each image manually first.
Once you have the imagery in the right space, then you should look into orthorectifying them - mosaicking comes later. In order to be able to orthorectify you will need to have specific camera model information (usually in the form of RPC or other camera model). You can do orthorectification as part of the mosaic dataset loading.
Lastly the single seamless image (mosaicking): If you chose to go with the mosaic dataset (need ArcGIS 10.0) then you can export the single image out of the mosaic dataset. If you don't have 10.0 or are not using the mosaic dataset, then you'll have to output each image for each processing step (georeferencing, orthorectification, ...) and then in the end use one of the mosaic geoprocessing tools to get your final output.
Finally a note: the jpg format has a 4GB limitation (size cannot exceed 65k x 65k), so if your final mosaic will be larger than that, you should consider a different format (such as tiff or img).
I hope this will get you started in the right direction.

Robert

[VarunShankar]

Hi again,

Thanks for your detailed reply, EinBerliner. I appreciate your help. I've stitched the images together to reduce the amount of work involved in georeferencing them (around 1200 pictures for each of the 19 passes made by the airplane)- now I merely need to georeference 19 images, instead of a bazillion.

I do have ArcGis 10. Based on your information, my process is now whittled down to:

1. Georeference each of the 19 photos (either manually or automatically).
2. Use ArcGis to orthorectify either each of the photos or the mosaic of the photos.
3. Export to jpeg/tiff.

I have photos shot from a thermal camera. I also have a file of the camera coefficients at each stage/pass- will this help in the georeferencing and orthorectification process? If so, could you tell me how? Is there perhaps some way to automatically georeference these using the camera coefficients that I have (this information isn't embedded into the photos, as far as I know, but I do have the information handy). I have some high level GPS information of the whole plot as well.

I'd appreciate your help, hopefully for one last time!

Thanks!

[RobertBerger]

Greetings.

Do you know what kind of camera was used on the airplane? If you do, I would suggest looking into loading the 1200 raw scenes into the mosaic dataset using the appropriate raster type if available (e.g. Applanix, ISAT, or MatchAT). This will automatically pull your camera metadata and you can provide an elevation model for automatic orthorectification. Also look for some notes at the bottom of the supported raster types list. If none of these work, then there is another way to load your ultracam data which might help you.
The problem with merging all your scenes first is that it is very hard after the merge to create an accurate camera model for the single dataset for orthorectification etc. Usually people orthorectify each scene first, and then in the last step merge their data together into a single dataset.
I hope this helps.

Robert