A commercial image processing system considered for triggering in future LHC experiments

Experiments at the future Large Hadron Collider (LHC) at CERN will be faced with an extraordinary problem of event selection in real time. Algorithms at different levels of locality are needed, among them local feature extraction algorithms that are typical in many image processing tasks. We report here about implementations of several such local algorithms on a commercial modular image processing system, for a decision frequency of 100 kHz (“second-level”). Such devices can keep up with the required decision frequency for suitable algorithms, for some algorithms custom hardware additions would be necessary.