These may occasionally give inconsistent results, such as blobs without features, but with skeletons. Ideally these would be made consistent. The latter method likely has faster performance, and if results are satisfactory could be implemented for the former blob tracking.
Creation of trajectories uses an adaptive otsu threshold https://github.com/ver228/tierpsy-tracker/blob/master/tierpsy/analysis/traj_create/getBlobTrajectories.py#L126-L132
Pharynx skeletonisation uses a blur followed by a global otsu+binary threshold https://github.com/ver228/tierpsy-tracker/blob/development/tierpsy/analysis/ske_pharynx/orient_pharynx.py#L29-L36
These may occasionally give inconsistent results, such as blobs without features, but with skeletons. Ideally these would be made consistent. The latter method likely has faster performance, and if results are satisfactory could be implemented for the former blob tracking.