Why Kruskal for track generation?

[AIBluefisher]

Hi, @mihaidusmanu

I'm curious about the manner of track generation. As is mentioned in your paper, you used kruskal to generate tracks with confidence that comes from descriptors similarity. It is different from the traditional track generation method which simply uses a union-find algorithm with limited size to generate tracks firstly and then filter ambiguous track elements. In my point of view, kruskal can generate more tracks but potentially shorter, union-find can generate longer but fewer tracks. Both methods may contain outliers and it's not clear which one is better. What do you think?