| Fire Detecting is an important using aspect of the development of Machine vision . Compare with the traditional technology of fire detecting, it is more intuitive, accurately and suitable for the network and the using in big space. The paper mainly focus on the fire that occurs on paper's daily producting and life . Analysis its flame image and find the method to recognite.Observate and statistic the flame image of fire that occurs on paper's daily producting and life, we can find its brighter and redder than other areas. So the paper studies the characters about the red degree of saturation and the average brightness of the fire flame . Find they both have certain laws .In order to reduce the disturbance of background information , we withdraw the flame by combining pictures and the method about red degree of saturation and average brightness. Since the fire flame's colour has so remarkable characteristic , then analysis the characters of flame color that obtained. Discovered the flame image's distributed situation in the relative two-dimensional space of color has obvious rule compare with the area that have no fire, The components of R and G both distribute in certain region.. Fit the boundary curve of the components of R and G . Because of the expansibility and beat with the time of the fire flame , the topic introduction two criterions. Computation the area of flame that continual fore images . We can find the area increases gradually. But stable fire flame is not the same. Remove the noise and disturbance and withdraw the outline of the flame . Track the outlines of continual four flame images that gotten . Then statistics the number of angles that lie in the outlines . We can find that the changing of the number have no certain rules . Although so is the stable fire flame , it's scope is smaller. If find the sutable threshold value, we can discriminate the stable fire flame. For a series of images , if they all satisfy the conditions above , fire occurs.The experiment result manifest that the method is easy and the realization is fast in the VB programe designed by myself. It can come up to 90 % above in recognition rate , and can fully meet paper's daily producting and life . |