论文部分内容阅读
采用传统单路传感光纤实现对信息特征的测量,常由于交叉敏感等不可控因素使测量数据出现异常值,导致信息分析偏差较大,识别准确度低。因此,提出了一种基于双相干谱、样本熵和奇异值分解(BSS)和反向传播神经网络(BPNNDS)算法的多路光纤入侵传感系统信息的特征提取与识别方法。针对含3路传感光纤的布里渊光时域反射(BOTDR)传感入侵检测系统,该方法利用BSS算法分别对不同入侵类型的多路信号进行特征提取;采用BPNN算法对不同入侵振动特征矢量进行分类;经Dempster Shafer(DS)证据理论算法对多路传感光纤的时空信息进行融合。数值分析与仿真实验结果表明,提出的信息提取方法可以有效提取出多路传感网络的信息特征,且使用BPNN-DS证据理论的多路信息融合方法能够准确识别多路入侵传感网络的信号类型,具有较高的准确度和可信度。
The traditional single sensing fiber to achieve the measurement of information characteristics, often due to cross-sensitivity and other uncontrollable factors make the measured data appear abnormal values, resulting in large deviation of information analysis, low recognition accuracy. Therefore, a method for feature extraction and identification of multi-channel optical fiber intrusion sensing system based on bispectrum, sample entropy, singular value decomposition (BSS) and back propagation neural network (BPNNDS) algorithm is proposed. For the BOTDR sensor intrusion detection system with 3-way sensing fiber, this method uses the BSS algorithm to separately extract the multi-channel signals with different intrusion types. BPNN algorithm is used to analyze the characteristics of different intrusion vibration characteristics Vector; classified by the Dempster Shafer (DS) evidence theory algorithm for the fusion of spatio-temporal information of the multi-sensing fiber. Numerical analysis and simulation results show that the proposed information extraction method can effectively extract the information characteristics of multi-sensor networks, and the multi-channel information fusion method using BPNN-DS evidence theory can accurately identify the signals of multi-channel intrusion sensor networks Type, with high accuracy and credibility.