Quantcast

Supervised Classification with ArcGIS

Posted by ForestryGIS on Tuesday, November 2, 2010

This post describes how to use the image classification tools in Image Analysis for Supervised Classification.

What You Will Need and Other Assumptions
• A satellite image that has been properly prepared for input to an image classification (e.g.,atmospheric correction, terrain correction, etc.)
• Aerial photographs and other ancillary data for the same area, as well as good local knowledge of the terrain, vegetation and soils for the area
• The Supervised Classification tools in Image Analysis are very limited as compared to those available in ERDAS Imagine

Overview of Steps
1. Load your satellite image to be classified into ArcMap’s Data Frame.
2. Create and load a polygon Shapefile for collecting your training signatures.
3. Format the Shapefile for signature collection and digitize training sites based on your classification scheme.
4. Specify the required parameters for the Supervised Classification tool and run your classification.
5. Inspect the results, modify existing signatures and collect new ones, and re-run your classification, if necessary.

Step-by-Step Example
1. Start ArcMap from your Desktop, or on the Windows Taskbar click Start | Programs | ArcGIS | ArcMap. Ensure the Image Analysis and Editor extensions are visible. If not, from ArcMap’s main menu select: 1) Tools | Extensions and enable Image Analysis; 2) View | Customize and enable Editor; and 3) View | Toolbars and enable Image Analysis and Editor.
2. Use the Add Data button on the main toolbar of ArcMap to add your satellite image to the Data Frame.
3. Click the ArcCatalog button in ArcMap (or start ArcCatalog from the Windows Taskbar) and navigate to your project directory.
4. In ArcCatalog, click File | New | Shapefile. In the Create New Shapefile dialog, name the new Shapefile in the Name field and change the Feature Type to Polygon.
5. Now click the Edit button beneath the Spatial Reference window in the Create New Shapefile dialog. Click the Import button in the Spatial Reference Properties dialog, navigate to the directory containing the satellite image you wish to classify, select it and click Add. Click Apply and OK in the Spatial Reference Properties dialog. You have just defined the Spatial Reference for your new Shapefile. Close ArcCatalog.
6. Use the Add Data button on the main toolbar of ArcMap to add your new Shapefile to the Data Frame.
7. Right-click on your Shapefile and select Open Attribute Table from the pop-up menu. Click the Options button in the attribute table and select Add Field.
8. In the Add Field dialog, name the new field ClassName (or something similar) and change the Type to Text. Click OK to add the new field to the table. Leave the attribute table open.
9. From the Editor toolbar, select Editor | Start Editing.
10. Select the Sketch Tool button from the Editor toolbar and digitize your first training site, and ensure the Target is set to your new Shapefile and the Task is set to Create New Feature (in the Editor toolbar). Note: use Single Left-clicks to add vertices to your polygons and a Double Left-click to close them.
11. After digitizing your first training site polygon, click in the ClassName field and type the name of your new training site polygon, based on your classification scheme.
12. Continue digitizing training site polygons and updating the ClassName field of the associated attribute table until you are satisfied that you’ve collected enough training sites for a complete and representative sample.
13. From the Editor toolbar, select Editor | Stop Editing and close the attribute table when finished collecting training sites. Click Yes when ArcMap asks if you want to save your edits. Alternatively, you can click Save Edits from the Editor pulldown menu before selecting Stop Editing. You are now ready to run a first iteration of your classification.
14. From the Image Analysis extension, select Image Analysis | Classification | Supervised. Specify the Input Image (your satellite image to be classified), the Signature Features (your Shapefile of training sites), and the Class Name Field (ClassName from your training site Shapefile). Select All Features for the Perform Classification Using option, select a Classification Rule (we recommend Maximum Likelihood, especially when running a Supervised Classification), and specify a location and filename for the Output Image (your new Supervised Classification). Click OK.
15. Inspect your results. Collect new signatures, modify existing ones, or delete bad signatures if necessary and rerun your classification until you achieve acceptable results.

Related Post



 
back to top