Speaker
Description
Time projection chambers generate large volumes of data that require substantial processing before physically relevant event information can be extracted. In the miniTPC detector installed at ELI-NP, each event is recorded in the U, V, and W readout views as a set of two-dimensional channel-versus-time frames. A dedicated 32-node Raspberry Pi cluster has been designed to support near-real-time analysis while remaining permanently associated with the detector. The system is installed in custom 19-inch rack-mounted enclosures and includes dedicated networking and power-distribution hardware. The analysis workflow uses computer-vision-based algorithms to process the U, V and W frames. These methods perform noise reduction, signal identification, feature extraction and correlation between the different readout views, providing the information required for three-dimensional event reconstruction. Because individual events can be processed independently, the analysis is well suited to parallel execution across multiple nodes. The cluster therefore provides a low-cost and scalable platform for local event reconstruction, data-quality monitoring, event visualization and the testing of new analysis methods. The present work describes the cluster architecture and the algorithms developed for the TPC data processing, together with the planned integration of the system into the detector data-analysis chain.