A Uav Vision System For Airborne Surveillance

The paper presents a machine vision system for aerial surveillance that can interpret and process data acquired by a UAV on-board infrared camera. System components include noise reduction, feature extraction, classification and decision-making. Decision-making is performed in terms of an alarm signal. The system has been configured for automatic fire-detection applications where the alarm is set off in case of fire identification. Real time tests have been performed and the system has been tested producing sets of real images. Finally, a genetic algorithm was used to automatically define some of the system's parameters.