Novel Methodologies for Pattern Recognition of Charged Particle Trajectories in the ATLAS Detector - PhDData

Access database of worldwide thesis




Novel Methodologies for Pattern Recognition of Charged Particle Trajectories in the ATLAS Detector

The thesis was published by Pitman Donaldson, Charlie, in July 2022, UCL (University College London).

Abstract:

By 2029, the Large Hadron Collider will enter its High Luminosity phase (HL- LHC) in order to achieve an unprecedented capacity for discovery. As this phase is entered, it is essential for many physics analyses that the efficiency of the re- construction of charged particle trajectories in the ATLAS detector is maintained. With levels of pile-up expected to reach = 200, the number of track candidates that must be processed will increase exponentially in the current pattern matching regime. In this thesis, a novel method for charged particle pattern recognition is developed based on the popular computer vision technique known as the Hough Transform (HT). Our method differs from previous attempts to use the HT for tracking in its data-driven choice of track parameterisation using Principal Component Analysis (PCA), and the division of the detector space in to very narrow tunnels known as sectors. This results in well-separated Hough images across the layers of the detector and relatively little noise from pile-up. Additionally, we show that the memory requirements for a pattern-based track finding algorithm can be reduced by approximately a factor of 5 through a two-stage compression process, without sacrificing any significant track finding efficiency. The new tracking algorithm is compared with an existing pattern matching algorithm, which consists of matching detector hits to a collection of pre-defined patterns of hits generated from simulated muon tracks. The performance of our algorithm is shown to achieve similar track finding efficiency while reducing the number of track candidates per event.



Read the last PhD tips