Setting up Training Sites Editing

Before you can perform a classification or an accuracy assessment, you must have ground-truth data. In remote sensing, ground truth refers to information collected on location. With ground-truth data, image data can be related to real attributes and materials on the ground. By collecting ground-truth data, remote-sensing data can be calibrated, which aids in the interpretation and analysis of what is being sensed.

With supervised classification, the ground-truth data acts as a training set, which is used by the learning algorithm to generate a classification model. In Object Analyst, you collect training samples for both the supervised classification and accuracy assessment in the same window.

Primarily, you will use on-screen interpretation in the viewer and select segments of various classes.

To set up Training Sites Editing

In the Object Analyst window, in the list under Operation, select Training Sites Editing.
The display is refreshed to show the settings for training sites editing.
Under Training Vector Layer, click Select, and then in the Layer Selection window, do the following:
1. In the File box, type or select the file you want.
  If the file you want does not appear in the list, click Browse, and then select the file you want.
2. In the Select column of the table, click the row of each layer you want.
  To select all the available layers, click Select All. A check mark appears in the Select column for each layer you select.
3. Click OK.
To modify the training sites, click Edit.
The Training Sites Editing window appears. For information about this window, see Using the Training Sites Editing window.
To import ground-truth points, under Ground-Truth Points, click Import.
The Import Ground-Truth Points window appears. For information about this window, see Importing ground-truth points.

After you finish setting up modifying the training sites, you can proceed to run a supervised or an unsupervised classification.