ArcGIS mosaic dataset: The next step in managing raster data

Some time ago, a client told me that he has about three hundred aerial images and fifteen digital elevation models (DEMs) of the same area. He wants to use these datasets to construct a single image that his technicians can use to view and analyze. Basically, he was looking for an option to store, manage, view, and query small to large collections of raster and image data in a rather simple way using ArcGIS, hey who doesn’t. :-)

Currently, the client and his team work with these raster datasets one at a time. For example, adding and checking to see which image covers their study area, one by one, which is, you know, very tedious.
 
Here is some good news: in ArcGIS 10.0, you may be glad to know that you can manage large amounts of raster datasets using a great new geodatabase data model called mosaic datasets. And yes, this functionality is available as part of the core.

What makes them different and great?
Here is the story – just like how one raster dataset stores data (photographic, elevation etc) as a continuous grid over a given area, mosaic datasets store many rasters over larger areas and present them seamlessly.

Not only large areas, but these mosaic datasets can bunch up individual datasets from different times – yes, I do mean temporal mosaics. If you have imagery of the same geographic area from several years, you can keep them all together in a single mosaic dataset, which could be used for tracking changes. And the datasets do not have to be adjoining or overlapping but can exist as unconnected or discontinuous datasets.

Also, some simple functionalities like hillshade, slope, aspect etc, which were available as separate tools in the Spatial Analyst toolbox are now a part of the mosaic dataset data model. Meaning, if you would like to be able to see hillshade of an area, you can now simply add that function to your mosaic dataset that contains your DEM(s); you don’t have to specifically run another tool and handle a different dataset for this purpose.

You can even Orthorectify your imagery, if you have the correct rational polynomial coefficient (RPC) information from your vendor.

Mosaic datasets also have advanced raster querying capabilities and processing functions and can also be used as a source for serving image services.
 
Doesn’t this sound great!? – a single dataset that extends the meaning and use of a raster dataset not only by blending their neighbors in space and time but also by offering derived products from the base rasters as functions to run and view on the fly. To top it off, these datasets carrying analytical results can be easily served to clients and users as seamlessly. Needless to say how many more steps and individual files does this save.

Contributed by: Sirisha Karamchedu
 

This entry was posted in Imagery and tagged , , , . Bookmark the permalink.

Leave a Reply

4 Comments

  1. rmvrtkn says:

    Will this dynamically work for non-georeferenced non-overlapping image tiles. For example, I take an image and slice it into 10 smaller images, if I add those to a new mosaic dataset, will it dynamically arrange it and correctly place each image tile.

  2. mharlow says:

    There is no automatic image-to-image registration.

  3. kpaybins says:

    sounds great, but are mosaic datasets the answer when I have 230 rasters similar to time-series in that they are each an identical extent and cellsize, but vary by data content. I would like to see all of these rasters in an ArcMap file all symbolized the same way- a classified symbology- 5 sets, a range of percentages, if that makes sense. I am looking for a way to apply a layer file to all of these at once, but when I have made a mosaic dataset (running ArcGIS 10.1) it is very time consuming.

    • Simon Woo says:

      A Mosaic Dataset is the the preferred way to manage your imagery data. It is fast, organized, and allows you to both manage and visualize your data.

      When you say that a mosaic dataset is “time consuming” what aspect is time consuming? A mosaic dataset should be relatively quick to create and display as well.