复杂噪声下相敏光时域反射仪传感信息提取与处理

Phase-sensitive optical time domain reflectometry(Φ-OTDR), can simultaneously detect and locate multiple weak disturbances at a quite long distance. Thus it has extensive applications in perimeter security, long pipe and large structure safety monitoring and so on. In China our group started the research of Φ-OTDR at the very beginning, and many key system parameters as sensing length and spatial resolution can be achieved at the top level in the world. However, when it is applied in real fields, it is found that the system's Nuisance Alarm Rate is quite high because it is also sensitive to the background noises. And the nonlinear characteristics of the sensing signal bring the hardness for the target recognition. The present signal processing methods in this area cannot meet the requirements of the field applications. Therefore, in this project our research aims to solve the detection and identification problems of Φ-OTDR systems in complicated noisy backgrounds. Firstly, according to the Time-Frequency(T-F) structure difference, it is proposed a kind of signal-noise and signal-signal separation solution based on the wavelet multiscale analysis and adaptive T-F decomposition. Secondly, the fractal theory is introduced for mining the inner regular features of the complex nonlinear signal, and a new detection and location method is then proposed based on the time-space fractal spectra and the energy distribution. The fractal characteristics of the multiscale signal components are also extracted and accompanied with the temporal statistics features and the T-F energy distribution spectra to fuse a complete but non-redundant feature vector, which is used for improving the typical signal's recognition rate. The research findings can not only meet the important state demand of safety monitoring well, but it can also be extended into many other interferencing fiber-optic sensing systems with high sensitivity. Thus there is universal significance to promote the development of information extraction and processing technology in the fiber-optic sensing area.

相敏光时域反射仪能够实现长距离、多点微弱扰动同时检测与定位，在周界安防、长输管道及大型结构安全监测等领域应用广泛。申请人所在项目组在国内率先开展该技术的研究，传感距离和空间分辨率已达到国际领先水平。但在应用中发现，其对环境中复杂噪声同时敏感，误报率较高，信号的非线性特征导致识别困难，现有信号处理方法无法满足实际应用要求。因此本项目以解决复杂噪声下光纤传感信号检测与识别的科学难题为目标开展创新研究：根据信号时频结构与分布差异，提出基于多尺度分析和自适应时频分解的信噪及信号物理分离方法；引入分形理论挖掘复杂非线性信号内在规律性特征，提出基于时空分形特征谱和能量谱的信号检测与定位方法；提出结合多尺度分形特征的多域特征融合方法，提高典型信号识别率。项目研究不仅满足国家重大安全监测需求，其成果能推广应用于其他干涉型光纤检测系统中，对于推动光纤传感信息提取与处理技术的发展具有普遍的重要意义。

本项目以解决复杂噪声下光纤传感信号检测与识别的科学难题为目标开展研究，分别解决了相敏光时域反射仪（Φ-OTDR）信噪及信号物理分离方法，基于多分析域的可分辨特征提取方法，以及基于多域特征融合的典型信号识别方法三大问题，并提出基于图像处理方法进行分布式光纤传感阵列信号检测与识别的新方法。最后，将研究成果全部进行应用转化。该研究满足国家重大安全监测需求，具有普遍的重要意义。通过本项目完成的具体内容包括：.（1）完成基于Φ-OTDR的埋地管道与埋地线缆安全监测的实际现场数据采集及火车、汽车等交通干扰、人为扰动及机械挖掘等典型事件数据库建立；.（2）对Φ-OTDR信号产生机制与非线性混叠机理进行了理论研究，提出基于多尺度分析的信噪及信号物理分离方法，检测信号信噪比提高 ~25dB，解决了Φ-OTDR传感信号和干扰噪声物理分离的难题；.（3）基于传感信号时频结构与分布差异特点，提取了传感信号的小波及小波包能量谱特征，以及针对传感信号随时间变化的非线性特征，提取了信号的奇异谱特征，研究了基于不同类型特征的不同事件识别结果； .（4）构建三层后向反馈（BP）神经网络，完成对典型背景信号、人为扰动信号及其他非平稳时变声音干扰信号的智能识别；提出基于主分量分析(PCA)的传感信号及特征降维方法，结合BP神经网络实现传感信号分类和识别；.（5）构建光时域反射仪时空信号二维演化谱，提出通过边缘检测及非线性滤波等图像处理方法进行事件目标检测、定位和识别的新方法；.（6）研究成果全部进行应用转化，基于研究的样机系统拓展了分布式光纤传感技术在国境线边防管控、埋地油气管道、输电线缆、供水管道等安全监测中的应用。.项目培养硕士研究生16名，共计发表SCI检索论文2篇，EI检索期刊论文2篇；参加国际会议交流10人次，发表EI检索会议论文10篇，其他会议论文3篇；申请中国发明专利11项，申请美国发明专利1项，获授权中国发明专利8项，登记计算机软件著作权共计9项。