分布式光纤安防检测系统的信号识别方法研究

发布时间：2018-12-23 08:00

【摘要】：随着社会的快速发展,通信电缆、输油管道、输气管道、高压电网以及军用、民用安防系统等基础设备、设施的安全监测等问题已日趋成为影响经济发展和社会稳定的重要因素。因此,针对各种危及电缆、管道安全及安防系统中非法入侵行为的及时发现和有效定位具有重要的研究价值和现实意义。本课题来源于与天津大学合作的国家973项目“光纤智能传感网实验平台关键技术及其应用的基础研究”课题,负责其中“恶劣环境对连续分布式传感网偏振及定位影响研究”子项目。本文以提高Mach-Zehnder干涉仪光纤安防系统扰动事件识别精确度为目标,采用小波多分辨率分析和神经网络等手段,对采集得到的分布式光纤传感安防系统的监测数据进行信号特征提取以及信号识别等研究,实现了对外界环境的下雨、敲击、攀爬等信号的识别,对提高安防监测系统中入侵行为的准确定位提供了研究基础。本文主要完成了以下几方面的工作：(1)采用小波变换,对分布式光纤传感安防系统的监测数据实现了多分辨率信息提取,通过小波能谱实现了信号特征的向量提取,获得了能较好反映各种安全事件的本质特征,对下雨、敲击、攀爬等信号实现了初步识别。(2)深入理解神经网络结构,将BP神经网络识别方法用于对安防检测信号类型的识别。设计了4种改进后的BP神经网络的MATLAB程序,并运用获得的样本数据的特征量对其分别进行训练,对训练结果进行比较分析,选取最佳BP网络训练方法。(3)对实验现场采集的监测数据进行了大量实验研究,证明了算法的有效性及可靠性,基于BP神经网络的识别率达到90%。
[Abstract]:With the rapid development of society, communication cables, oil pipelines, gas pipelines, high-voltage power grids, military and civilian security systems and other basic equipment, Safety monitoring of facilities has become an important factor affecting economic development and social stability. Therefore, it is of great research value and practical significance to detect and locate the illegal intrusion behavior in the cable, pipeline safety and security system. This topic comes from the national 973 project "key Technology and basic Research on the Application of Optical Fiber Intelligent Sensor Network experiment platform", which is in cooperation with Tianjin University. Responsible for the "adverse environment on the continuous distributed sensor network polarization and positioning research" subproject. In order to improve the accuracy of disturbance event identification in Mach-Zehnder interferometer optical fiber security system, wavelet multi-resolution analysis and neural network are used in this paper. The signal feature extraction and signal recognition of the monitoring data of distributed optical fiber sensing security system are studied, and the recognition of rain, knock, climbing and other signals in the outside environment is realized. It provides a research basis for improving the accurate location of intrusion behavior in security monitoring system. The main work of this paper is as follows: (1) Multiresolution information is extracted from the monitoring data of distributed optical fiber sensor security system by wavelet transform, and vector extraction of signal features is realized by wavelet spectrum. The essential characteristics of various security events are obtained, and the initial recognition of the signals such as rain, knocking, climbing and so on is achieved. (2) the neural network structure is deeply understood. The BP neural network recognition method is used to identify the type of security detection signal. Four improved MATLAB programs of BP neural network are designed, and the training results are compared and analyzed by using the characteristic quantity of the obtained sample data. The optimal BP network training method is selected. (3) A large number of experimental studies are carried out on the monitoring data collected from the experiment site, which proves the validity and reliability of the algorithm. The recognition rate based on BP neural network reaches 90 points.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP274;TN253

【参考文献】