Felix: playing around with ITS raw data to 'validate' the staggered readout&reconstruction.

Gabriele:

• Changed algorithm
• Moved to algorithm inspired by Run2 implementation
• Basically it is a 3D histogramming, but "Raytracing vertexer" sounds cooler
• Define a 3d voxelisation around the beamline
•  
• Define tracklets as lines parametrized as p(t) = p₀+tu
  where p₀ is cluster on innermost ITS layer, u is direction of the line
• For each track
  • Calculate t_enter and t_exit for the voxelisation
  • Sample the line at uniform t_step between this range
  • Each time the line steps into a voxel, update the count

• Select the local maximums that surpass a threshold
• Then fit the lines that passes through the vertex candidates
• Below video where low and high mult vertices are detected, sliding through z
• video
  
  
• Result with only candidates finding, without fitting:
  Precision: 0.9924, Recall: 0.8667, F1: 0.9253
  Total true vertices: 150
  Matched vertices: 130
  Missed true vertices: 20
  Total fitted vertices: 131
  Fake vertices: 1
  Mean residuals (x,y,z): [0.00014556 0.00058302 0.00027437]
  Std  residuals (x,y,z): [0.00432429 0.00439    0.00784673]
• Next steps:
  • Just select highest mult vertex
  • do an histogram along that z to find lower mult vertices