METHOD FOR ABNORMALITY DETECTION USING IMAGE

PURPOSE:To provide a method for abnormality detection using an image and its device capable of stably detecting an abnormal event such as the spouting of flame, steam, black smoke, etc., through image processing and stably detecting abnormality without being affected by the effect of transient disturbance such as the passage of an operator and sunshine changes. CONSTITUTION:An image pickup device picks up plural input color images Ci (i=1, 2..., K, K+1,...n) of an object at proper time intervals, converts the respective input color images Ci into monochromatic images Mi in order, and obtains difference images M2 as differences between monochromatic images Mk+1 and Mk, thereby detecting abnormality of the object from the difference cumulative image SIGMAM2 obtained by cumulatively adding the difference images M2. Further, the difference images M2 are binarized with a binarization threshold value for difference images to obtain a difference binarized image M3, and a fluctuating area of an input color image Ci is extracted from the difference binary image M3, and, a color deciding process for the fluctuating area is performed to decide which of flame, steam, and black smoke the extracted abnormality is.