New to Meteorology ideas in storm Identification and tracking

Where in the world is Lak?

Thanks to Don MacGorman, Will Agent

The common approach

Objects identified based on a threshold
All pixels above threshold

Problem: threshold is global

Same threshold does not work for initiating vs.

Example of threshold problem

Problem: Association is final

Association takes only two frames into account
Bad decisions

Example of association problem

Premise …

Try to avoid hard decisions
Use locally adaptive thresholds to identify

Enhanced Watershed Transform

Start from local peak
Grow till specified size is reached
In

EWT Example

Multiple Hypotheses Tracking (MHT)

MHT is based on two useful algorithms:
Hungarian Method

MHT
In practice, will lead to combinatorial explosion
So, prune to keep around

EWT and MHT in QC of Az-Shear

Azimuthal Shear a very noisy

Rotation Tracks Cleanup

Summary

Can avoid/postpone hard decisions in tracking
Use locally adaptive thresholds to identify