This article describes an experiment to evaluate an autonomous automatic tunnel incident detection system. The system uses EVA (Equipment Visionique Autonome). Details about the configuration of the EVA system are provided. The method used to monitor the accuracy of the system is also presented. The criteria used to ascertain the accuracy of the system are: a) detection rate; b) false alarm rate; c) average freqeuency of false alarms; and d) average duration of a false alarm. The aid algorithm was validated over an 1872 hour period. During this time 843 incidents were identified. The detection rate was 100%, and the false alarm rate was one false alarm every 84 hours. A second experiment was carried out to ascertain hardware disfunction. This was carried out over a two month period with 243 incidents detected.
Abstract