摘要
为了快速准确地对图片中的火焰进行分割检测,消除灯光等疑似火焰区域的干扰,提出基于RGB和HSV颜色空间的火焰检测模型并结合CNN提取火焰轮廓。通过观察RGB三分量在颜色直方图中的分布特征,制定新的约束条件,并结合HSV颜色空间,消除复杂背景对其进行的干扰,最后利用CNN对提取出火焰轮廓进行分类检测,进一步消除对灯光等疑似火焰区域的误检。实验结果表明,该算法可以准确地检测出火焰区域,降低了误检率。
In order to detect the flame quickly and accurately and eliminate the interference of suspected flame area,a flame detection model based on RGB and HSV color space and CNN(convolution neural networks)are proposed to extract the flame contour.By observing the distribution characteristics of RGB three components in the color histogram,new constraints are formulated,and combined with HSV color space,the interference of complex background is eliminated.Finally,CNN is used to extract flame contour to further eliminate false detection of suspected flame areas.The experimental results show that the algorithm can detect the flame accurately and reduce the false detection rate.
作者
李莹
李忠
李海洋
孙可可
韩成果
Li Ying;Li Zhong;Li Haiyang;Sun Keke(Institute of Disaster Prevention,Emergency Management College,Hebei 065201,China)
出处
《计算机时代》
2019年第12期67-70,共4页
Computer Era
关键词
火焰检测
卷积神经网络
颜色空间
轮廓特征
flame detection
convolutional neural networks
color space
contour feature
