基于改进梯度边缘特征的早期火焰识别方法  

Early flame recognition method based on improved gradient edge feature

作　　者：刘小虎[1] 欧阳继能 卜乐平[1] LIU Xiao-hu;OUYANG Ji-neng;BU Le-ping(College of Electrical Eng in ecring, Naval Engin eeri ng Univer sity, Hubei Wuhan 430033, China)

机构地区：[1]海军工程大学电气工程学院,湖北武汉430033

出　　处：《消防科学与技术》2019年第2期250-252,共3页Fire Science and Technology

基　　金：2018年度海军工程大学科研发展基金自主立项项目(425317Q033)

摘　　要：在视频火灾探测中,基于颜色特征变化的火灾识别模型适应性差、对火焰倒影等在颜色模型方面与火焰图像很类似的外界干扰分辨率差。利用火焰燃烧时在空气中的状态与周围环境的颜色阶跃性,在颜色分量模型的基础上,对火焰图像边缘进行粗定位和精定位,使用改进的Soble边缘检测算法,选取颜色B分量的边缘梯度作为判断火焰和干扰图像的依据,构建了一种基于图像边缘梯度的火焰识别模型。实验证明:颜色分量梯度模型能够排除大量的常见干扰源,准确、快速地对火灾进行识别,对大空间室内火灾探测具有一定的使用价值。In video fire detection , the fire recognition model based on color feature change has poor adaptability and poor resolution of exter nol interfere nee, which is similar to flame image in color model. Based on lhe color component model, the edge of flame image is roughly and precisely posilioned. Then the edge gradient of color B component is selected as the basis of judging flame and interfer encc image by using lhe improved Soble edge detection algorithm. A flame recognition model based on image edge gradient is constructed. The experiments show that the color component gradient model proposed in (his paper can eliminate a large number of corrr mon interference sources, identify fire accurately and quickly, and has certai n application value for large space in door fire detection.

关 键 词：边缘梯度 火焰颜色模型 火灾探测 

分 类 号：X924.4[环境科学与工程—安全科学] TP277.2[自动化与计算机技术—检测技术与自动化装置]