基于多空间颜色特征的火灾检测

Fire Detection Based on Multi-space Color Feature

针对传统的烟雾和温度火灾探测器仅适用于空气流动性较低的室内空间,并且在发出警报前,烟雾和温度都需要达到一定的阈值等问题,本文提出了一种基于多空间颜色特征的火灾检测算法。首先,对采集到的图像进行预处理,利用高斯滤波和直方图均衡化进行图像去噪与图像增强。然后对图像进行混合高斯背景建模,实现动态目标的检测。对动态目标区域提取RGB和HSI空间相联合的颜色特征,根据多空间的颜色特征获取火焰区域的掩模图像,实现对视频图像中火焰区域的检测。并对掩模图像作开运算操作,去除图像中的噪点和毛刺,平滑不平坦边缘,最后对火焰区域进行连通域检测,并绘制矩形框,从而实现火灾的检测。

In view of the problems that traditional smoke and temperature fire detectors are only suitable for indoor spaces with low air mobility,and the smoke and temperature need to reach a certain threshold before an alarm is issued,this paper proposes a fire detection algorithm based on multiple space color features.First,preprocess the collected image.Use Gaussian filtering and histogram equalization for image denoising and image enhancement.Then the image is modeled with Gaussian mixture background to realize dynamic target detection.Extract the color features combined with RGB and HSI space for the dynamic target area,and obtain the mask image of the flame area according to the on multiple space color features to realize the detection of the flame area in the video image.The mask image is opened and operated to remove the noise and burrs in the image,smooth the uneven edges.Finally,detect the connected domain of the flame area and draw a rectangular frame to realize fire detection.