Available with Image Analyst license.
Available with Spatial Analyst license.
In supervised image classification, you need to train the classifier to assign pixels or objects to a given class using training samples. The class categories are determined by your classification schema, and the training samples can be generated using the Training Samples Manager pane. Tools in the Training Samples Manager pane allow you to create training samples for each class category in your schema and provide information about the number and size of samples to help you improve the accuracy of your classification model.
The Training Samples Manager is found in the Classification Tools drop-down menu in the Image Classification group on the Imagery tab. Select the raster dataset you want to classify in the Contents pane to display the Imagery tab, and be sure you are working in a 2D map. The Classification Tools menu is unavailable if the active map is a 3D scene, or if the highlighted image is not a multiband image.
The Training Samples Manager is also displayed in the classification wizard workflow and operates in the same manner as described below.
Manage the classification schema
When you open the Training Samples Manager pane, you can see the schema management section at the top, with the default classification schema from the 2011 National Land Cover Database (NLCD2011). If your classification schema is only moderately different from the NLCD2011 schema, you can add or remove individual classes from this schema and save it as a new schema. Right-click any of the classes to add classes or edit the class properties, such as the name, value, color, and alias. You can also load an existing schema or create a new schema and save it.
Schema edits should be made before collecting training samples. If edits are made to a class after samples in another class have been collected, temporarily switch to a different sketch tool to preserve edits to the class.
Create a training sample by drawing a polygon around pixels or objects in the raster.
Create a training sample by drawing a circle around pixels or objects in the raster.
Create a training sample by drawing a freehand shape around pixels or objects in the raster.
Create a training sample by selecting a segment from a segmented layer. This option is only available if there is a segmented layer in the Contents pane. Activate the Segment Picker by highlighting the segmented layer in the Contents pane, and then select the layer from the Segment Picker drop-down list.
Create a new classification schema. Right-click the New Schema title and click Add New Class to begin creating class categories.
Select a classification schema option.
Save changes to the schema.
Save a new copy of the schema.
Add a class category to the schema. Select the name of the schema first to create a new parent class at the highest level. Select the name of an existing class to create a subclass.
Remove the selected class or subclass category from the schema.
Create training samples
Complete the following steps to create the training samples:
- Load the classification schema you want to use in the schema manager at the top of the Training Samples Manager pane using the Classification Schema button . Add or remove class categories if you want to make modifications. Save any changes you make to the schema.
- Select the class that you want to collect training samples for from the list of classes in the schema manager.
- Click one of the sketch tools or use the segment picker to begin collecting training samples.
- To use the Segment Picker, the segmented image must be loaded into the Contents pane. Click the drop-down arrow to select the segmented layer that you want to collect training samples from.
- Click a segment in the map to add it as a training sample.
- Using a sketch tool, delineate the image feature representing the class on the map. Collect a representative number of training samples for each class in your schema.
- Add, delete, and organize your training samples using the tools in the bottom section of the pane. Once you are satisfied with the training samples, click the Save button to save your results.
Manage the training samples
The bottom section of the pane displays and manages the training samples you have collected for each class. Collect representative sites, or training samples, for each land cover class in the image. A training sample has location information (polygon) and an associated land cover class. The image classification algorithm uses the training samples, saved as a feature class, to identify the land cover classes in the entire image.
You can view and manage training samples by adding, grouping, or removing them. When you select a training sample, it is selected on the map. Double-click a training sample in the table to zoom to it on the map.
Open an existing training samples feature class.
Save edits made to the current training samples feature class.
Save the current training samples as a new feature class.
Collapse multiple training samples into a single multipart training sample. This can be useful if you want to see the total number of samples for each class category and assess the distribution of sample sizes for each class, or if you want to delete a large group of training samples at once.
Expand a multipart training sample into its individual component features. Each training sample becomes its own polygon.
Duplicate the selected record for the slice in the multidimensional raster layer that is currently being displayed. This is useful for generating training samples for multiple time slices if the sample class has not changed over time.
This tool is only activated if the raster being sampled is a multidimensional raster layer.
Delete the selected training samples.
The training samples table in this portion of the Training Samples Manager pane lists the number of samples and the percentage of pixels representing each class. If you used the Segment Picker to collect your training samples, the number of samples is the number of segments you selected to define the class. This is important to keep in mind when you use a statistical classifier such as Maximum Likelihood, because the number of segments represents the total number of samples. For example, if eight segments were collected as training samples for a class, it may not be a statistically significant number of samples for reliable classification. However, if you collected the same training samples as pixels, your training sample can be represented by hundreds or thousands of pixels, which is a statistically significant number of samples. The number and percentage of training samples is less important when using the nonparametric machine-learning classifiers such as Random Trees and Support Vector Machine.
Collect time information in samples
The Training Samples Manager supports the collection of time information for each training sample when the raster being sampled is a multidimensional raster or multidimensional mosaic dataset. If the Multidimensional Info has been built for a time series of raster images, training samples can be collected for each slice in the dataset, and the slice's time information will be automatically generated in the training sample attributes.
- Click the StdTime drop-down menu in the Current Display Slice group on the Multidimensional tab to display the slice for which you want to collect samples.
For more information, see Multidimensional tab.
- In the Training Samples Manager pane, select the class that you want to collect training samples for from the list of classes in the schema manager.
- Click one of the sketch tools and delineate the image feature representing the class on the map.
- If a sample persists multiple times in the multidimensional raster layer, you can duplicate it across multiple time slices.
- Select the sample in the table.
- Use the controls in the Current Display Slice group to switch the display to the new time slice.
- Click the Duplicate for Current Display Slice button in the Training Samples Manager to create a copy of the sample with a time value matching that of the current slice.
You can collapse multiple samples from the same class and the same time into a single record, and then duplicate the collapsed record to copy all the training samples to a new time slice.