Forest fires represent a constant threat to ecological systems, infrastructure and human lives. The only effective way to minimize damage caused by forest fires is their early detection and fast reaction, apart from preventive measures. Great efforts are therefore made to achieve early forest fire detection, which is traditionally based on human surveillance. Traditional human surveillance is direct human observation by observers located on monitoring spots. Technologically more advanced system is the use of video monitoring systems, and the most advanced approach is automatic surveillance based on automatic early forest fire detection. Today there are several different approaches but the most feasible one is system based on video cameras sensible in visible spectra. In almost every country which encounters high risk of forest fires at least one such system was developed. In Croatia there is also such system called Intelligent Forest Fire Monitoring System (iForestFire® ). The system was successfully implemented in various Croatian regions, National and Nature Parks. This paper describes the main concept and the main idea of iForestFire system and particularly explains its “ intelligent” parts – design based on the observer network theory, system multi-agent architecture and detection algorithm based on advanced image processing and analysis procedures.

Računarstvo