topic Re: Iso Unsupervised Classification in Data Management Questions

Iso Unsupervised Classification

GeospatialTechnology — Mon, 18 Jun 2012 14:01:31 GMT

Hello,

Just a couple of questions about the Iso Unsupervised Classification function in the Image Classification Toolbar to make sure I understand what I'm doing.

For Minimal Class Size ArcGIS Desktop help states:

"??? To provide the sufficient statistics necessary to generate a signature file for a future classification, each cluster should contain enough cells to accurately represent the cluster. The value entered for the minimum class size should be approximately 10 times larger than the number of layers in the input raster bands"

How do I find out the number of layers in the input raster bands? Is this simply the number of bands. So for example for a RGB aerial photograph with 3 bands the minimal class size would be 30?

For sample interval ArcGIS Desktop help states:
"
??? The value entered for the sample interval indicates one cell out of every n-by-n block of cells is used in the cluster calculations"

So for a sample interval of 10 on cell out of every 10 by 10 block of cells in used in the cluster calculations?

Re: Iso Unsupervised Classification

XuguangWang — Mon, 18 Jun 2012 17:04:50 GMT

Yes, your understanding is correct for both statements.

Re: Iso Unsupervised Classification

GeospatialTechnology — Tue, 19 Jun 2012 08:20:46 GMT

Thanks Xuguang, is there way to determine the optimum sample interval size for a particular raster dataset?

Re: Iso Unsupervised Classification

XuguangWang — Tue, 19 Jun 2012 22:57:17 GMT

The Sample Interval parameter is there to accelerate the execution of IsoCluster process.

A large Sample Interval tends to skip small features on the image. If you are trying to identify small features on the image, use a smaller Sample Interval.

The default value (10) supplied for Sample Interval on the tool dialog should be good for most use cases.

Thanks,
Xuguang

Re: Iso Unsupervised Classification

GeospatialTechnology — Wed, 20 Jun 2012 07:55:11 GMT

Thanks Xuguang, yes im trying to identify small features. I'll play around with some different values and see which gives the best results. Thanks for your help in this.