基于YCbCr颜色空间和BP神经网络的火焰图像检测

A Flame Image Detection Method Based on YCbCr Color Space and BP Neural Network

根据火灾燃烧时会产生红色光和蓝色光的火焰的特性,采用具有亮度和红、蓝2种颜色信息的YCbCr彩色空间提取火焰特征.采用BP神经网络建立火灾检测模型,YCbCr所提取的亮度分量和红、蓝2个颜色分量作为神经网络的输入.利用100组样本图像训练神经网络模型,通过反向传播来改变神经网络的权值和阈值,从而减小误差.利用10组火灾样本图像和10组干扰样本图像验证模型的有效性.

According to the characteristic that,when fire burns,it produces red and blue flames,YCbCr color space with brightness and red and blue color information is used to extract the flame characteristics.BP neural network is used to establish the fire detection model.The brightness,red and blue extracted by YCbCr are used as the input of the neural network.The neural network model is trained with 100 sample images.The weight and threshold of the neural network are changed by back propagation to reduce the error.The validity of the model is verified by using 10 groups of fire sample images and 10 groups of interference sample images.