A C + + object-oriented toolkit for track finding with k-dimensional hits

Abstract A library is described for the recognition of tracks in a set of hits. The hits are assumed to be k -dimensional points ( k - d ), with k ≥ 1 , of which a subset can be grouped into tracks by using short-range correlations. A connection graph between the hits is created by sorting the hits first in k - d space using one of the developed, fast, k -space containers. The track-finding algorithm considers any connection between two hits as a possible track seed and grows these seeds into longer track segments using a modified depth-first search of the connection graph. All hit-acceptance decisions are called via abstract methods of an acceptance criterion class which isolates the library from the application's hit and track model. An application is tuned for a particular tracking environment by creating a concrete implementation for the hit and track acceptance calculations. The implementer is free to trade tracking time for acceptance complexity (influencing efficiency) depending on the requirements of the particular application. Results for simulated data show that the track finding is both efficient and fast even for high noise environments.