DBSCAN based detection algorithm

[damb]

Background: scdetect-cc allows for multi-stream template based detector configurations. Within the context of a detector each stream is cross-correlated individually and the local maxima are forwarded to the so-called linker. The linker's task is to associate the local maxima under consideration of certain criteria (i.e. the template, "arrivalOffsetThreshold", "minArrivals"). Once these criteria are met, there is a detection which is used for further processing.

While linking local maxima for detectors with a few streams is computationally feasible, it becomes more expensive with a growing number of streams to be associated.

Tasks:

• Determine how many streams can be handled with the current implementation
• Evaluate the usage of the DBSCAN algorithm

Resources:

• https://en.wikipedia.org/wiki/DBSCAN
• https://www.researchgate.net/publication/323317120_How_the_DBSCAN_Algorithm_works_A_simple_introduction_using_sketches
• https://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.html
• https://github.com/uofuseismo/mflib

[damb]

@mmesim, further input is welcome.