Abstract: In response to the requirements of flame detection in different fire scenarios, this study establishes a fire detection model with fire characteristics and investigates the mechanisms of multi-band red and ultraviolet (UV) sensors for flame detection, multi-feature data fusion algorithms, and fire decision-making methods. Ultraviolet photodetectors in the wavelength range of 185 to 260 nanometers and infrared photodetectors in the wavelengths of 3.6, 4.4， 4.8 μm are selected for flame detection, and flame characteristic signal data information is collected. By employing multi-signal data fusion and feature parameter correlation algorithms, a mathematical model for fire detection based on multi-band fire characteristics is established. Fire occurrence scenarios are simulated to train the optimal configuration of relevant feature parameters. The research results demonstrate that multi-band flame detection using UV light in the range of 185 to 260 nanometers and infrared light at wavelengths of 3.6, 4.4, 4.8 μm can meet the requirements of the national standard GB 15631-2008， Special fire detectors. The recognition and decision-making capabilities for flames are improved, and there were no false alarms or missed detections in the experiments.

Key words: UV photoelectric sensor, infrared photoelectric sensor, multi band flame detection