Random sample consensus: Difference between revisions

Revision as of 23:27, 20 February 2023

Revision as of 00:00, 22 February 2023

Line 24:

# All data are then tested against the fitted model. All the data points (of the original data) that fit the estimated model well, according to some model-specific [[loss function]], are called the ''consensus set'' (i.e., the set of inliers for the model).

# The estimated model is reasonably good if sufficiently many data points have been classified as a part of the consensus set.

# Afterwards, theThe model may be improved by re-estimating it withby using all the members of the consensus set. The fitting quality as a measure of how well the model fits to the consensus set will be used to sharpen the model fitting as iterations goes on (e.g., by setting this measure as the fitting quality criteria at the next iteration).

To converge to a sufficiently good model parameter set, this procedure is repeated a fixed number of times, each time producing either the rejection of a model because too few points are a part of the consensus set, or a refined model with a consensus set size larger than the previous consensus set.