Rule-guided identification of cosmic-ray patterns in PLASTEX

Some techniques devised in the computer science fields of pattern recognition and expert systems are being applied to the interpretation of EAS responses in the PLASTEX experiment. An attempt is made to codity in a set of rules the expertise of trained researchers who are able to recognize and classify different hit patterns even in the presence of noisy background, and in spite of imperfections in the detector response. The patterns expected to be useful include, but are not limited to, track patterns. The software described here, as a progress report, automatically finds patterns corresponding to isolated tracks, and patterns composed of tracks that connect with each other in a layer of dense material. Tracks are reconstructed iteratively, using both geometrical and internal consistency arguments. Multihits, geometrical inefficiency, mechanical imperfection and track bending are taken into account. Data preprocessing is applied to multihits according to a theoretical model of the particle-detector interaction, and the presence of tracks in dead detector areas is inferred by probabilistic arguments. Finally, particle scattering is allowed for using a variable-size tolerance cone during the track reconstruction procedure. Examples are given of application of the method to recognition of nearly straight tracks, such as those expected to be made by high-energy muons, and of more complicated patterns made of connected straight tracks. © 1992 Società Italiana di Fisica