With supervised classification, you must provide a training set with examples for each class. The algorithm "learns" from the training sample to create a classification model. In the training phase, you select samples and create a training set from which the classifier can learn.
From previous field work, image or map analysis, personal experience, or combination thereof as an analyst you are familiar with the identity and location of samples belonging to classes. You locate the specific sites (samples) in the data that represent examples of known classes. Various supervised-classification algorithms can then learn from these examples and build a model to predict class membership of data for which membership is unknown.
When creating a training set, an important consideration is that it must provide a representative description of each class; that is, it must have completeness.
In Object Analyst, you can create classes, select segments polygons, and then link them to appropriate classes through onscreen interpretation. The class information for selected segments is stored in a Training attribute field. Color information for each class is saved in the file metadata and used post-classification to create a thematic map. You can create various classes and assign a color to each. Later, when you render the classified layer, each class is displayed with its assigned color.
© PCI Geomatics Enterprises, Inc.®, 2026. All rights reserved.