基于图像处理的明火火灾探测研究

A fire-detection method based on image processing technique

基于数字图像处理技术,实时采集火灾现场帧彩色图像,实时计算图像数据在空间分布上所包含的纹理特征信息——能量(CON)、熵(ENT)、相关(COR)、惯性矩(ASM),并提取火焰的形状,获得火焰特征参数——面积、周长、圆形度、重心.通过融合处理连续帧图像数据,量化得到纹理信息表示的现场环境状况和火焰特征参数表示的火焰形状及其不稳定性特征。这些特征参数在火灾发生时出现了明显的变化,呈现出一定的规律,利用这些规律可以在现代船舶上实现火灾图像探测报警。

This paper, based on image processing technique, deals with the collection of images of a fire on the spot and the calculation of the data of image distributed over an area on real time. The data contains characters of texture - CON, ENT, COR, ASM and also takes the shapes of fire and obtains the parameters of fire flame-area, perimeter, degree of roundness, center of gravity. Through handling successive data of each image, the information of texture concerning the field environment and the parameters of fire flame clearly demonstrates the shapes of fire flame and the unstable characters. When a fire happens, the parameters of fire will change obviously and regularly. The regular change can be used for the fire detection, which will play an important role in morden ships.