The Study On Early Flame Detection Based On Infrared Thermal Imaging

Fire disaster has always been regarded as the frequently occurred disaster with huge collapsing force.It always threatens people's life,destroys material wealth and ecological balance.As a result,severe negative influences have been brought to the development of human society and economic construction.At the present,it is the most practical and effective methods to detect and rescue at the very beginning of the occurrence of fire disaster for maximally reducing damages in case of the occurrence of fire disaster.Traditional fire detection technologies including sensing smoke,light sensing and detectors are limited to the using occasions.In the meantime,it is hard to detect fire disaster in the early stage.Along with the incessant development of various scientific technologies in recent years,including pattern recognition,computer image processing technology and image sensor,the image-type-based fire detection technology developed on this has already been popular among people and gradually being used to various occasions,in which there is one-dimensional temperature information more available for infrared thermal imaging technology based on pattern information.It has precious value for fire detection in the early stage and also is helpful for fire warning and trend analysis,based on which the early-stage flame detection algorithm based on characteristics of infrared thermal imaging is proposed and applied into the fire detection system by adopting image processing technology and combining infrared thermal imaging.The specific work in this paper is as below:Firstly,the research background and status of image-type flame detection technology is analyzed and it is compared into the traditional fire detection technology.In addition,the advantages and disadvantages of image type flame detection technology are summed up and the status of infrared thermal imaging fire detection technology is also analyzed in this dissertation.By analyzing the characteristics of infrared thermal imaging,a new image segmentation method is proposed in this dissertation.This method can be adopted to accurately extract the flame goal under area with high luminance and small goals in fire.Meanwhile,an algorithm module and morphological operation process applicable for flame detection is selected according to experimental comparison.Through analysis of flame dynamic characteristics according to the characteristic of low image resolution of thermal imaging,three calculation methods for three dynamic characteristics are proposed;one group of multi-dimension eigenvectors and flame characteristic model are built in this dissertation.Then,the identification of suspicious goals can be realized through C4.5 decision tree classification method.According to the experimental result,it shows that this algorithm is simple and has good classification ability.Lastly,by applying the fire detection algorithm into flame identification system and utilizing the temperature and image characteristic under thermal imaging based on the abovementioned work,the detection of early-stage fire is realized and the experimental result that this system is applied into various flame detections is given out.