Overlapping orthophotos in a Mosaic Dataset

[schignel]

Hi all, I've a number of orthophotos covering a forest in flight lines that slightly overlap. I first attempted to mosaic them all together into a single TIFF with image analysis, but it's such a big file that I'm now trying to put them in a mosaic dataset.

I have a couple of questions:

1. The only way i can get a decent mosaic without cracks and spaces between the images is to use the "max" mosaic operator, which works fine when I'm zoomed in, but the overviews I built don't seem to want to mosaic in the same way--they are all chopped up when I'm even a bit zoomed out. Is there a way to have the overviews also mosaic nicely?


2. Which would be the best mosaic method to use in this case? I'm using North-West as a default and it works ok, but would closest to center make more sense?


3. I have sets of images from 5 sequential years. Would it be best to include them all in a single mosaic dataset and give them each a Year field and then choose the mosaic method "by attribute?" OR would it be wiser to make a different Mosaic Dataset for each year?

Thanks!
Steve

[Hornbydd]

Stephen,

Don't know about questions 1 or 2 but if it were me for 3 I would have a separate mosaic for each year as that is easiest to understand for others. So unless there is some sort of improved performance issue by stacking them and filtering them temporally simple is best and probably easier to document with metadata?

Duncan

[jgamble]

Hey Steve,

1. I was in a similar spot a few months back where using the max operator introduced black pixels or gaps into my overviews. However, the mosaic looked perfect when looking at scales that displayed raw images. Our issue was caused by the pyramids built for the TIFFs. Removing the pyramids on all of our images and rebuilding the overviews fixed the problem. Not sure if you are using pyramids, but if you are, try rebuliding a mosaic dataset using a small subset of TIFFs without pyramids and then build the overviews.

2. I think it depends on how the imagery was taken, and how you need each image to be placed within the mosaic. For the mosaic dataset I created using the max operator, I needed to display my images in the North-West fashion. You could also create a sort order yourself and use the 'by attribute' method.

3. Ditto to what Duncan said above about creating a separate mosaic dataset for each year. An additional question would be do you need to mosaic all images spanning the years into a single mosaic? If so, you could look into creating a derrived mosaic dataset and add the contents of the 5 separate mosaic datasets into the derrived md.


Cheers
Jon

[schignel]

Thanks for the responses guys. I will definitely be making a separate mosaic dataset for each year then.

OK, so I removed all of the pyramids (built them to level "0"), loaded them into the mosaic dataset and built the overviews but they look very much the same as before. Any other suggestions?


I have had so much trouble trying and trying to get mosaic datasets to work. They seem like they have so much potential but they just never seem to do what I want them to...

[voldune]

Hi Steve,

For #1, I would suggest rebuilding your footprints using the Build Footprints too, and make sure you uncheck Skip Overviews. You can find this same information in the Problems with black overlap while mosaicking satelite images forum thread.

Patrick

[jgamble]

... And if rebuliding footprints does not work - can you post up a screen capture of what the mosaic currently looks like? May be able to offer some further suggestions or identify what you may be doing wrong.

[schignel]

Ok, rebuilding the footprints helped a great deal-- thanks voldune for that wonderful link. The black stuff has gone, and I've figured out how to set the default mosaic operator within arcCatalog. "Max" is the current setting, although maybe that's no longer the best since I'm not worrying about the black pixels anymore?

Now my only problem is that some of my overviews are some how overlapping each other-- creating double features that don't exist. Here's an example of a road that looks like 2 roads...



When I zoom in to the actual rasters everything looks great, it's only the overviews.

[jgamble]

Hmmm ... I don't think max is the operator you are looking for now.

How many different raw images are we looking at here? Is it possible that there are two overlapping images that have that road appearing in them, and these images are not aligning correctly?

Have you tried removing the overview images and re-building the overviews?

[schignel]

Ok, I decided to start over with a fresh MD and did a test with 3 rasters. I added them in, built footprints, and built the overviews using the blend mosaic operator, which is what ESRI suggestions for continuous data (which I think a series of ortho photos counts as). Here's how they're looking:



Notice there are still some doubling of features between the top and the middle images? Do I just need to be satisfied that the mosaic isn't going to be perfect? Does anybody know what the best mosaic operator is for digital ortho photo flight lines?


Also, once I figure this out, I want to use model builder to automate this whole process; Is there a way to change the default mosaic operator (to "blend" for example) of my mosaic dataset before running the build overviews tool in the model, so that my overviews come out blended?

Thanks for the continued help through this!
Steve

[jgamble]

Steve,

are you sure your two images align correctly? When looking at the raw images (not overviews), do you see this duplication? Based on the attached image, I would say they do not align and thats why you see the duplication ...


And as for the model builder - I'm not 100% sure how you would accomplish this in model builder.

[schignel]

Yep, you're right. Looking just at the raw images they still overlap. Bad rectification I guess.

Thanks again for all the help.