Classification of a Block Model

You can modify classification settings, then classify blocks. Classification is written to the model classification variable and is used to assign confidence classes such as Measured, Indicated, and Inferred.

Classification Logic

• Measured: a block is measured if it meets the selected settings.
• Indicated: generally based on an extrapolation from measured support, according to the selected method.
• Inferred: blocks not classified as measured or indicated are classified as inferred.

Access from the Block Models List

Use the Estimation menu:

• Classification Settings…: opens the classification parameter dialog.
• Classify: runs classification using the saved settings on the selected block model(s).
• Clear Classification…: removes selected class assignments from selected model(s).
• Copy Classification…: copies classification settings from one model to other model(s).

Classification Parameters

The classification settings dialog includes four main tabs: Method, Search, Envelopes, and Add/Edit Class.

Method

Choose one of the available methods:

• User defined script: define the classification method with your own script.
• Classical measured, indicated and inferred evaluation: standard measured/indicated/inferred evaluation using an indicated-size factor.
• Classical By Ellipsoid (One defined class by ellipsoid): uses one class pass per ellipsoid/class setup.
• By Filling Ellipsoid (Use of an algorithm centered on the composites): uses a composite-centered filling approach.

Method Details

Classical measured, indicated and inferred evaluation

This is the standard workflow for a measured/indicated/inferred result.

• A measured zone is evaluated first using your search setup.
• The indicated zone is then derived from measured support using the indicated-size factor.
• Blocks not meeting measured or indicated conditions remain inferred.

Use this method when you want a simple and fast classification flow controlled mainly by one scaling factor.

For this method, the indicated-size factor is set in this tab (valid range: 1 to 100).

Classical By Ellipsoid (One defined class by ellipsoid)

This method applies classification in multiple passes.

• Each pass uses its own search/ellipsoid settings.
• Each pass is associated with a target class.
• Pass order matters: the class assignment follows the sequence of your configured passes.

Note that a block will only be classified by a better classification, so you must start with infered at pass 1 and finish with measured at pass 3

By Filling Ellipsoid (Use of an algorithm centered on the composites)

This method is also pass-based, but centered on composite distribution and local filling behavior.

• Classification is driven by how composites populate the search volume.
• Each pass can target a specific class.
• This method is suited to workflows where continuity and composite coverage are key classification drivers.

Use this method when you want classification behavior that is more directly tied to composite-centered support than a simple size-factor extrapolation. Reduce spotted dog effect and more logical classification than the Classical by ellipsoid method. This is the recommended method to use.

Search

Select the ellipsoid and search parameters used to classify blocks.

This tab defines the search behavior for classification passes, including composite-based search settings when required by the selected method.

Envelopes

Assign optional envelopes to limit:

• block screening, and/or
• composite screening.

In many workflows this is not required if composite sets are already prepared with domain limits.

Add/Edit Class

Add or modify class labels and class indices used by the classification model.

This is typically only needed when your project uses custom class definitions.

Smoothing

You can enable Smoothing using second pass moving window average.

This post-processing step uses a selected ellipsoid to apply a moving-window average to classification results. It is useful to reduce isolated patterns (“spotted” effects) and produce a more continuous classification map.

Practical Notes

• Classification is applied to all currently selected block models.
• A valid composites set must be defined in search settings before running classification.
• If the selected composites set is missing, classification is interrupted and an error is shown.
• For methods using multiple passes, each pass can classify a different class according to your search/pass setup.
• If smoothing is enabled, choose a smoothing ellipsoid before running Classify.
• Default class structures often include an Unclassified class in addition to measured/indicated/inferred classes.