This paper proposes a feature-based multi-sensor fire detector operating on ordinary video and long wave infrared (LWIR) thermal images. The detector automatically extracts hot objects from the thermal images by dynamic background subtraction and histogram-based segmentation. Analogously, moving objects are extracted from the ordinary video by intensity-based dynamic background subtraction. These hot and moving objects are then further analyzed using a set of flame features which focus on the distinctive geometric, temporal and spatial disorder characteristics of flame regions. By combining the probabilities of these fast retrievable visual and thermal features, we are able to detect the fire at an early stage. Experiments with video and LWIR sequences of lire and non-fire real case scenarios show good results in id indicate that multi-sensor fire analysis is very promising.