Refereed full papers (journals, book chapters, international conferences)

1993

We have taken a set of well-understood algorithms and combined them in a commercial application: train-spotting. An autonomous train-spotting system requires expertise in active vision, tracking and model-based recognition. Results are demonstrated from real image sequences showing that BARRY can indeed perform its task satisfactorily for many hours without the sustenance that human spotters require (flasks of hot tea, sandwiches etc.). The maximum train velocity is at present limited, but we demonstrate theoretically that the system could in principle operate with trains travelling at relativistic speeds, with obvious application to future space-based spotting.