Representative spectrum

Takes the spectrum from numbers of objects to make representative spectrum. The pixels can be chosen using a number of different algorithms, explained below.

Parameters

Method

• Evenly Spread

  • The pixels are evenly spread across the images/objects.

• Random

  • The pixels are randomly spread across the images/objects.

• Random(Gaussian)

  • The pixels are randomly spread with a larger concentration of pixels in the center of the images/objects.

• Spectral Evenly

  • The pixels are distributed using PCA score and spread evenly sorted by sum of squares for all components

• Spectral Binning

  • The pixels are distributed using PCA to get a spread containing a variety of spectrum for the objects.

• Spectral Space Filling (Very slow)

  • The pixels are distributed using PCA and iterated to find the spectrum containing the best variety to describe the spectrum of the objects.

• Spectral Clustering (k-means)

  • The pixels are distributed using k-means clustering to partition the pixels into k number of clusters and select pixels for the partitioned clusters.

Number

Number of objects to create, .

Unique

✅ Adds only unique spectrum

⬜ Can have the same spectrum from different representative spectrum pixels.

Dimensions

Size of the objects created from the segmentation

1x1 will create objects 1 pixel in size.

Clusters

Only applicable for Spectral Clustering (k-means)

Specify number of clusters () to partition the observations into.

Applies to

Only visible when applicable

When Applies to is used only objects from the selected segmentation will be used for the next segmentation on the analyse tree.

This is denoted by the dashed line from the Object node to the segmentation which only is applied to a subset of all applicable segmentation.

Link

Only visible when applicable

Link output objects from two or more segmentations to top segmentation. Descriptors can then be added to the common object output and will be calculated for objects from all segmentations.

The segmentations must be at same level to be available for linking.