ΦOTDR光纤入侵检测识别理论基础研究

发布时间：2018-06-14 08:21

本文选题：光纤预警系统 + 检测识别算法　；参考：《北方工业大学》2017年硕士论文

【摘要】：本文围绕相位光时域反射(ΦOTDR)体制下的光纤入侵检测识别算法理论基础研究展开。主要通过设计出ΦOTDR光纤预警系统的检测识别算法实现有害入侵信号的定位和类型识别。首先,在检测算法研究方面,本文深入研究了典型的恒虚警率(CFAR)检测方法并设计出了新的空间维度CFAR检测方法。并且在时间维度进行非参数检验方法设计,实现了无害干扰的去除。其次,在识别算法研究方面,本文研究了基音周期(PP),占空比(DC),过零率(ZCR)等特征的提取方法,为有效识别入侵类型提供基础。最后,通过运用经典的视觉注意架构,使整体算法结构合理。鉴于此,本文设计出了基于视觉注意架构的光纤预警检测识别算法,并进行了该算法的实验验证。首先,对光纤预警系统的检测算法进行研究。在空间维度检测中,本文提出一种新的自适应背景匀质性CFAR检测方法,该方法能够在保证检测性能的同时能够尽可能的减少算法的时间消耗,保证噪声被剔除;在时间维度检测中,本文提出了针对无害干扰信号的频繁程度选择进行序贯似然比(SPRT)或K-S检测的处理,保证无害干扰信号被剔除。其次,对光纤预警系统的识别算法进行研究。发现机械入侵信号存在PP,人工挖掘信号存在较小的DC,过车信号存在较小的ZCR,并且实验证明不同类型的实测数据能够提取到相应特征。最后将检测识别算法融合到视觉注意架构中,将算法分为数据驱动和任务驱动两部分。并且实验证明数据驱动更能够有效地减小后续处理的数据量,任务驱动能够使有害入侵类型识别率明显提升。
[Abstract]:This paper focuses on the theoretical research of optical fiber intrusion detection algorithm based on phase optical time domain reflection (桅 OTDR). The detection and recognition algorithm of 桅 OTDR optical fiber early warning system is designed to locate and identify harmful intrusion signals. Firstly, in the aspect of detection algorithm, this paper deeply studies the typical CFAR detection method and designs a new spatial dimension CFAR detection method. The nonparametric test method is designed in time dimension to remove harmless interference. Secondly, in the research of recognition algorithm, this paper studies the extraction methods of pitch period, duty cycle and ZCRs, which provide the basis for the effective identification of intrusion types. Finally, by using the classical visual attention architecture, the overall algorithm structure is reasonable. In view of this, an optical fiber early warning detection and recognition algorithm based on visual attention architecture is designed and verified by experiments. Firstly, the detection algorithm of optical fiber early warning system is studied. In spatial dimension detection, a new adaptive background homogeneity CFAR detection method is proposed. This method can not only guarantee the detection performance, but also reduce the time consumption of the algorithm as much as possible, and ensure that the noise is eliminated. In time dimension detection, the sequential likelihood ratio (SPRT) or K-S detection is proposed to select the frequency of harmless interference signal to ensure that the harmless interference signal is eliminated. Secondly, the recognition algorithm of optical fiber early warning system is studied. It is found that there are PPPs in mechanical intrusion signals, small DCs in manual mining signals and small ZCRs in vehicle passing signals. The experimental results show that different types of measured data can extract the corresponding features. Finally, the detection and recognition algorithm is integrated into visual attention architecture, and the algorithm is divided into two parts: data driven and task driven. Experimental results show that data drive can reduce the amount of data in subsequent processing more effectively, and task driven can significantly improve the recognition rate of harmful intrusion types.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP277

【参考文献】