摘要
火灾火焰特征的提取是图像型火灾火焰探测中的关键问题,该文提出了一种基于RBF神经网络算法的图像型火灾识别的方法,首先运用中值滤波的方法对图像进行预处理,然后提取疑似火焰区域的颜色决策、面积变化率和闪烁频率特征,最后将获取的特征值作为RBF神经网络的输入量,通过神经网络的训练获取这些特征参数包含的火灾信息,从而判断是否为火灾图像,仿真实验结果证明了算法的准确性和实时性。
The key problem of image fire detection is the fire flame feature extraction.A image fire detection method research is proposed in this paper based on RBF neural network.The first use of the method about median filter for image preprocessing,Then the extraction of suspected flame area’s characteristics about color decision,area change rate and flicker frequency.Finally using the acquired characteristic values as input of RBF neural network,through the training of the neural network to obtain the fire information form the characteristic pa rameter contains.In order to determine whether the fire image,the simulation results show that the accuracy of the algorithm and re al-time.
出处
《电脑知识与技术(过刊)》
2011年第11X期7988-7990,共3页
Computer Knowledge and Technology