2) Boundary Classification

Description:

Boundary classification refers to training a machine learning algorithm to predict boundary probabilities for lines obtained through image segmentation. Training data is required to train the classifier, which line represents a desired outline before it can predict boundary probabilities for unseen lines.

Input:

• image segmentation lines without attributes .shp
• RGB orthoimage raster .tif
• DSM orthoimage raster .tif (optional)

Output:

• image segmentation lines with boundary likelihood as attribute per line .shp

Steps:

The prediction can be realized through RF Classification or CNN Classification classification.