abstract |
A method for detecting flame and smoke using spatial analysis of video input ( 40 ) provided by a video detector. The video input ( 40 ) consists of a number of individual frames, wherein analysis is performed on each individual frame to detect and outline regions that may contain smoke or flame ( 42 ). Based on the defined outline or boundary of each detected region, spatial features associated with the region are extracted ( 52 ), such as perimeter/area and surface area/volume. The extracted spatial features are related to one another ( 54 ) to determine the likelihood that the region contains smoke or flame. Extracted spatial features may be related to one another using a power law relationship that provides an indication of the turbulence associated with a bounded region, wherein turbulence is a characteristic of both flame and smoke. |