基于多参数融合和RBF神经网络的泥石流预报

摘要

In order to improve the debris geological disaster predicting accuracy,and solve serious problems arising from single monitoring method,which leads to high false positive rate and false negative rate,multi-parameter fusion and RBF neural network are used for comprehensively analyzing multi-parameters collected by multi-sensors,which makes prediction results more scientific for the decision-making departments.Input of RBF neural network is multi-sensor parameters,kernel function of hidden layer is Gaussian function,and output is debris flow occurrence probability.By training the RBF neural network,the prediction model can be obtained,when input data is collected presently,so as to make the right prediction decision.Simulation and experimental results show that the method proposed can improve the accuracy of predicting debris flow,and improved prediction results can provide more scientific basis for decision-making departments,which would protect people's lives and property.%为提高泥石流地质灾害预报准确率,改善传统泥石流监测和预报中存在的监测方法单一而导致的误报和漏报的问题,采用多参数融合和RBF神经网络相结合实现对多个参数的综合分析,得到比较准确的泥石流灾害发生概率,RBF神经网络的输入为多传感器采集到的多参数,隐含层采用高斯核函数,输出层为泥石流发生概率.通过训练RBF神经网络,获得泥石流预报模型,实时采集的多参数输入到训练好的模型,可以计算出泥石流发生概率.仿真和实验验证表明,该方法能够有效提高泥石流灾害预报准确率,可以为决策部门提供更加科学的预报结果.