摘要
根据火灾发生时着火面积逐渐增大、火焰边缘抖动等特征 ,提出了一种基于 BP网络的火灾图像探测系统。着重讨论了利用具有二次收敛性的 L evenberg- Marquardt学习算法进行 BP网络的权值优化 ,与标准 BP算法相比 ,用 L evenberg- Mar-quardt学习算法训练的
A fire image detection system based on BP neural network is presented in this paper according to the characteristics of fire,such as flaming area becoming larger and larger,fire edge wobbling,etc.The focus of our discussion is the optimization of BP neural network adopting Levenberg- Marquardt algorithm that has the prop- erty of quadratic convergence.Compared with the standard BP algorithm,the BP neural network trained by Lev- enberg- Marquardt algorithm has the quicker convergence rate and better reliability.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2001年第4期437-440,共4页
Chinese Journal of Scientific Instrument