Automated Shape Filtering

One of the keys to successful shape detection is being able to filter the outputs from the algorithm to match your desired outcomes. Targeting provides the following filters:

• Area: filter identified grains by their calculated area (in µm²). Note that the calculated area may be affected by any scaling inaccuracies in your region.

• Major Axis Length: Set bounds for the longest axis through the identified grain boundaries using this filter (see Fig. 9)

• Minor Axis Length: Set bounds for the shortest axis through the identified grain boundary (see Fig. 9)

• Circularity: Set bounds for the circularity index of each identified grain (Fig. 9). Circularity is defined as [Example Source].

• Convexity: Set bounds for the convexity index of each identified grain. Convexity is defined as the area of the shape divided by the area of a convex hull of the show. A “convex hull” in this case is a shape enclosing the grain without any inward dents (i.e. it is perfectly convex). The commonly used example is wrapping a shape in a rubber band. The rubber band will skip over indents in the shape. If the object is very near circular without any dents, the convexity will be near 1. (Fig. 9)

• Distance: Exclude targets within this distance (in µm) of a grain. This can be useful if sub-grains have been listed multiple times in TIMA datasets.

Fig. 9 The minor and major axes of an example grain outline (top). Examples of low, medium and high convexity (middle) and circularity (bottom).

Contours define grain boundaries but if grains have inclusions, previous ablation pits etc, there may be interior contours within the grain boundary. Parent contours contain at least one child contour. Child contours have at least one outer (parent) contour. Outermost contours have no containing contour. The “Types” dropdown menu provides options for filtering contours (and hence target grains) by their hierarchical relationship to other contours.