基于亮度与火焰区域边缘颜色分布的火焰检测*

为改善目前的火焰检测方法对环境适应能力不强的情况，提出一种基于亮度与火焰区域边缘颜色分布的火焰检测方法。主要采用二值重构、形态学算法以及边界追踪计算火焰区域边缘颜色分布矩阵，对得到的颜色分布矩阵进行主成分分析（PCA），并用PCA中协方差特征值分量约束BP神经网络的输入向量，从而准确进行了火焰检测。实验结果表明，此算法计算简单，能准确识别多种背景下的火焰图像。

Fire detection based on brightness and color distribution of fire edge regions

In order to improve the situation which most fire detection methods are not strong to adapt to a variety of environment, this paper proposed a new fire detection method based on the brightness and color distribution of the fire edge regions. It calculated the color distribution characteristic matrix of the fire edge regions using reconstruction, binary morphological and boundary-tracing algorithms. Then processed the color distribution matrix with PCA algorithm. And bound BP neural network input vectors by covariance Eigen value of PCA components. Through that, conducted fire detection accurately. The experiment shows that this algorithm is simple, and can identify fire image accurately against a variety of backgrounds.