Artificial timeseries to explore some more interesting object linking scenarios.

This artificial timeseries builds on the Definiens-produced training examples to provide some more complex object linking scenarios:

objects with no overlap between frames
an object that buds and then fuses back together
objects that pass by each other too closely to trivially separate
an object that is not visible for one of the time frames.

The example ruleset illustrates the following ideas:

generation of levels after object detection to visualise linking on demand (run deactivated process at end)
itterative evaluation class based linking
forwards and backwards linking to handle both budding and fusing
elimination of crossing point by detection of multiple in and out links
examination of incoming and outgoing link angles to find and eliminate kinks in link chain caused by incorrect linking
linking across temporal gaps by use of eval class to find chain starts that related to preceeding chain ends.

I hope someone finds this useful!

The ideas in this example were formulated with my Tessella colleagues during a training session run by Florian Leiss.

• Example analysis.dcp(314 KB )
• Frames.zip(103 KB )
• sshot.png(2.3 KB )
• For downloading this file, please login or sign up for a new account.