PatForward

PatForward is a pattern recognition algorithm designed to find long tracks. It starts from Velo seeds, adds the corresponding T station track segments and any potentially matching TT hits. It is described in the following notes:

• The Forward Tracking: Algorithm and Performance Studies
• The forward tracking, an optical model method

The idea behind PatForward is that hit positions in T stations can be predicted for all z positions if track parameters (x,y,tx,ty) in the Velo and either the track charge and momentum or the position of a single hit in the T stations are known. For each Velo seed, one can then project T station hits into a reference plane and look for clusters to find the hits that belong to a Velo seed. (The track model used in this algorithm thus serves as Hough transform, and the reference plane is the Hough space used.)

The algorithm works in two steps, collecting x hits and thus fixing track parameters in xz projection comes first, then, stereo hits are collected in much the same way. Once the full set of 3D parameters has been determined, the algorithms picks up nearby TT hits and finishes with a ghost and clone killing stage.

-- ManuelSchiller - 13 Aug 2008