Unsupervised Learning for Pixel Mask Clustering and Cluster Tracking in LHCb's Velopix Sensor Calibration
M.W. Majewski, P. Radoń, T. Szumlak
AGH University of Science and Technology in Krakow, 30 Mickiewicza Avenue, 30-059 Kraków, Poland
Full Text PDF
The silicon vertex detector is one of the core elements of the LHCb spectrometer. Its upgrade version features an innovative pixel sensor. The readout chip branches from the Medipix family of dedicated pixel ASICs. One of its operational challenges with the future data taking at the Large Hadron Collider will be the ability to detect faulty pixels and monitor them. In this work, we propose an innovative method for clustering faulty pixels and tracking their evolution in time. We compare the two methods of clustering (DBSCAN and OPTICS) and their influence on the proposed tracking method using a simulated dataset of masked pixels.

DOI:10.12693/APhysPolA.142.418
topics: unsupervised learning, VeloPix, tracking, particle detectors