|
|
Clustering hits and predictions in data from TimePix3 detectors
Čelko, Tomáš ; Mráz, František (advisor) ; Holan, Tomáš (referee)
Hybrid pixel detectors like Timepix3 and Timepix4 detect individual pixels hit by particles. For further analysis, individual hits from such sensors need to be grouped into spatially and temporally coinciding groups called clusters. While state-of-the-art Timepix3 detectors generate up to 80 Mio hits per second, the next generation, Timepix4, will provide data rates of up to 640 Mio hits, which is far beyond the current capabilities of the real-time clustering algorithms, processing at roughly 3 MHits/s. We explore the options for accelerating the clustering process, focusing on its real-time application. We developed a tool that utilizes multicore CPUs to speed up the clustering. Despite the interdependence of different data subsets, we achieve a speed-up scaling with the number of used cores. Further, we exploited options to reduce the computational demands of the clustering by determining radiation field parameters from raw (unclustered) data features and automatically initiating further clustering if these data show signs of interesting events. This further accelerates the clustering while also reducing storage space requirements. The proposed methods were validated and benchmarked using real-world and simulated datasets.
|
|
|
Support for annotating and classifying particles detected by Timepix3
Čelko, Tomáš ; Mráz, František (advisor) ; Holan, Tomáš (referee)
Hybrid pixel detectors like Timepix3 can capture gigabytes of data on various particles in a second. However, in such measurements, a vast majority of these particles represent already well-known particles. Distinguishing between the types of particles is the first step in searching for extraordinary particles. It is a non-trivial task often done by physicists. Source data consists of clusters that are groups of pixels of the detector hit by a particle or its secondary particles when the particle decays. Manual processing of the data to such an extent is inefficient. We created a set of tools for visualizing clusters, computing properties of clusters, filtering clusters based on their properties, and training neural network classifiers. Trained classifiers can be merged into a tree structure, offering a better utilization of unevenly distributed types of clusters. Based on simulated labeled data, we trained multiple classifiers and evaluated their performance on the test dataset of clusters.
|
guest ::
