Reverting to the old TITAN overlap neighbourhood has been useful, but it is still desirable to have a meaningful physical basis for this size of the neighbourhood applied. Tests could use the feature elongation information along with some idea of the feature size to only select smaller features, which are more prone to be lost during tracking.
This is adding a bit more complexity to the method than the previous version, but using a neighbourhood can always be disabled in the config by setting the overlap_nbhood parameter to 0 (this should be mentioned more clearly in the docs)
Reverting to the old TITAN overlap neighbourhood has been useful, but it is still desirable to have a meaningful physical basis for this size of the neighbourhood applied. Tests could use the feature elongation information along with some idea of the feature size to only select smaller features, which are more prone to be lost during tracking.
This is adding a bit more complexity to the method than the previous version, but using a neighbourhood can always be disabled in the config by setting the overlap_nbhood parameter to 0 (this should be mentioned more clearly in the docs)