基于颜色特征的火焰检测新算法

New algorithm of flame detection based on color features

针对当前计算机视觉火焰检测算法无法实现高准确率和低误报率的问题,提出了一种基于颜色模型的候选火焰图像元素分类算法。该算法首先对RGB模型各通道求均值获得新的图像样本,再利用YCbCr颜色空间建立火焰图像元素分类模型,通过设计YCbCr颜色模型新规则来减少由于图像亮度发生变化而产生的干扰,火焰像素检测率得到显著提高,能够较为准确地识别火焰,有较高的实用价值。

Using computer vision techniques to achieve early fire detection is a hot research topic.In this paper,a candidate flame pixel classification algorithm is proposed based on color model.First,the features of the flame region in RGB and YCbCr color models are introduced.Then YCbCr model is used to build flame pixel detection model.In addition by converting the rules based on RGB color model into corresponding rules in YCbCr color model,new rules in YCbCr color model are developed,which can further alleviate the harmful effects of changing illumination.The method can be used for real-time fire detection.