Abstract

By using the ratio of the G component to the R component G/R, or the ratio of the B component to the R component B/R, from a television monitor image, and referring to a conversion table for converting to distribution temperature, a surface temperature of a flame is found for each pixel. An area in which luminance signals included in an image of a surveillance area exceed a prescribed level is sampled as a flame outline, and at least the distribution temperature of a sampled flame region is detected, and distance to a radiant energy source is measured. The amount of radiant energy from the radiant energy source is estimated by computation, based on the sampled flame area and its distribution temperature and the distance to the source of radiant energy. The system is provided with a fire image recognition section for recognizing a fire from radiant energy itself and from change in radiant energy over time and outputting fire detection information and an intruder entity recognition section which, from monitor images from the same camera, recognizes an intruder entity and outputs burglary detection information. Processing modes are provided enabling fire and/or intruder entity image processing to be selected. If abnormality detection information is received using fire image or intruder entity recognition processing, a pre-alarm is outputted in the central monitor room, and also the monitor image in the central monitor room is switched to the image of the location where the abnormality has been detected.