single record

Algorithm/Service Info

Algorithm or Service name
Blurred Cluster
Author or Maintainer
Mike Wallbank
one line description
2D cluster reconstruction technique which specialises in clustering hits from shower deposits by first applying a weighted Gaussian smearing to the hit map in order to more accurately distribute the charge and form more complete clusters.


The algorithm takes 2D hits (previously found by other reconstruction techniques, e.g. gaushit) and outputs 2D cluster objects (saved as recob::Clusters) into the event.  Ideally, each cluster contains all hits from deposits left by a single particle as it traverses the TPC.  The method specialises in shower-like particle deposits, particularly in the separation of separate but nearby showers (e.g. reconstruction of pi0 decay).

The reconstruction proceeds in the following general steps:

  • An initial analysis of the hits first provides information such as direction and size of the showers which is used later on.  The ‘width’ of the hits in time (i.e. width of fitted Gaussian) is also used.  [Currently, this assumes all showers travel in a similar direction — this will be generalised soon.]
  • The hit map is convolved with a 2D Gaussian kernel in order to smear the deposited charge.  The kernel selected uses information about the hits obtained in the previous stage.  The charge from each hit is distributed according to the kernel, which also changes on a hit by hit basis depending on the ‘hit width’.
  • This process introduced so-called ‘fake hits’, areas of charge not corresponding to charge read out from the wire planes.  It thus encourages a more realistic spread of charge, preferentially along the direction of the shower, and ensures fuller clusters are formed.
  • Clustering proceeds on the ‘blurred hits’ using a simple nearest-neighbour approach up until a given charge threshold.
  • Finally, the fake hits are removed and the initial hits from each cluster are saved as a recob::Cluster object with associated recob::Hits.

See the document attached to this entry below for a more complete description, with example event displays.  This also contains information on configuring the module and the free parameters available for tuning, along with selected parameters for the DUNE 35t experiment and recommended procedures for tuning further.  [MW: this document is currently very out of date!  In the process of rewriting…]

location in code
2D hit reconstruction, other
code analysis done
improved code released

Author Info