基于MET地震检波器的周界入侵防御系统

发布时间：2018-04-15 20:53

本文选题：防入侵 + 目标识别　；参考：《吉林大学》2015年硕士论文

【摘要】：目前常用的周界入侵防御系统的探测传感器主要有声、地震动、被动红外、磁、电、视频等传感器，但是由于很多物理量容易受到外界环境的影响，导致基于该类传感器系统的性能难以达到较高的要求，地震检波器具有受多普勒效应和环境影响最小的优势，因此被广泛应用于周界入侵防御系统中。但目前市场上常用的地震检波器自然频率较高且灵敏度较低，探测距离较近的同时会丢失部分目标识别需要的低频成分，因此有必要寻找一种高灵敏度的超低频地震检波器。MET地震检波器具有灵敏度高、自然频率可达1Hz的优势，因此本文开发基于MET地震检波器的周界入侵防御系统，同时选取目前最常用的传统地震检波器——JF-20DX-10Hz作为对比探测传感器。本文的主要研究内容为：第一，研究了地震动信号的基本理论，分析了目标运动时地震波的产生及传播机理，对包含地震波主要能量的瑞雷波的特性进行详细阐述，为后续数据处理做理论基础。第二，对MET地震检波器的结构和工作原理进行了介绍，并设计一系列实验对其实际性能进行测试，分别为自身噪声测试、灵敏度测试、相频特性测试，自身噪声测试使用目前地震领域的国际标准测试其噪声加速度功率谱密度和速度功率谱密度，灵敏度和相频特性测试使用超低频振动台对其进行扫频测试得到其灵敏度及相频特性，通过上述三个实验证明MET地震检波器的实际性能符合其标定指标。第三，，研究了基于小波包组合BP神经网络的目标识别算法。首先使用过零分析和峰度分析在时域上对地震动信号进行分析，使用功率谱分析在频域上对信号进行分析；然后对小波分析的原理进行介绍，并介绍了常用的小波函数，学习小波去噪算法的原理，学习小波分析的改进算法——小波包分析在特征提取方面的应用；接着对BP神经网络的原理进行介绍，学习其网络结构、训练过程及其在目标识别方面的应用；最后确立小波包组合BP神经网络的目标识别算法模型，确定小波包特征提取算法，构建BP神经网络并对网络的训练流程进行详细阐述。第四，开发周界入侵防御系统软件，软件使用NI LabVIEW进行开发，编写良好的人机交互界面实现采集数据的显示、保存及处理。使用NI公司的TDMS文件格式进行数据保存；使用LabVIEW调用MATLAB的方式实现目标信号的特征提取和BP神经网络的训练及识别，将识别结果以文字的方式直接显示在系统软件主界面上。为了进一步提高目标识别的准确率，增加离线分析模块对采集数据进行进一步分析。外场实验结果表明，相比JF-20DX-10Hz地震检波器，基于MET地震检波器的周界入侵防御系统具有更高的识别精度和识别距离。
[Abstract]:At present, the detection sensors of the perimeter intrusion Prevention system are mainly acoustic, ground motion, passive infrared, magnetic, electric, video and other sensors. However, because many physical quantities are easily affected by the external environment,Because the performance of the sensor system based on this kind of sensor system is difficult to meet the higher requirements, the geophone has the advantage of least influence by Doppler effect and environment, so it is widely used in perimeter intrusion prevention system.However, the seismic geophone, which is commonly used in the market, has high natural frequency and low sensitivity, and it will lose some of the low-frequency components needed for target recognition while the detection distance is close.Therefore, it is necessary to find a highly sensitive ultra-low frequency geophone. Met geophone has the advantages of high sensitivity and natural frequency up to 1Hz. Therefore, a perimeter intrusion prevention system based on MET geophone is developed in this paper.At the same time, JF-20DX-10 Hz, the most commonly used geophone, is chosen as the contrast detection sensor.The main contents of this paper are as follows:Firstly, the basic theory of ground motion signal is studied, the generation and propagation mechanism of seismic wave are analyzed, and the characteristics of Rayleigh wave, which contains the main energy of seismic wave, are described in detail, which is the theoretical basis for the subsequent data processing.Secondly, the structure and working principle of MET seismograph are introduced, and a series of experiments are designed to test its actual performance, which are noise test, sensitivity test and phase frequency characteristic test.Self noise testing uses current international standards in the seismic field to test the noise acceleration power spectral density and the velocity power spectrum density.The sensitivity and phase frequency characteristics of MET geophone are measured by scanning frequency with ultra-low frequency vibration table. The results show that the actual performance of MET seismograph accords with its calibration index.Thirdly, the target recognition algorithm based on wavelet packet combination BP neural network is studied.At first, zero-crossing analysis and kurtosis analysis are used to analyze the ground motion signal in time domain, and power spectrum analysis is used to analyze the signal in frequency domain, then the principle of wavelet analysis is introduced, and the commonly used wavelet function is introduced.Learning the principle of wavelet denoising algorithm, learning the application of wavelet packet analysis in feature extraction, then introducing the principle of BP neural network, learning its network structure.The training process and its application in target recognition. Finally, the model of target recognition algorithm based on wavelet packet combination BP neural network is established, the feature extraction algorithm of wavelet packet is determined, the BP neural network is constructed and the training flow of the network is described in detail.Fourthly, the software of perimeter intrusion prevention system is developed. The software is developed with NI LabVIEW, and a good man-machine interface is written to display, save and process the collected data.The TDMS file format of NI company is used to save the data, the feature extraction of target signal and the training and recognition of BP neural network are realized by LabVIEW calling MATLAB, and the recognition result is displayed directly on the main interface of the system software in the way of text.In order to further improve the accuracy of target recognition, an off-line analysis module is added to further analyze the collected data.The results of field experiments show that the perimeter intrusion prevention system based on MET geophone has higher recognition precision and distance than JF-20DX-10Hz geophone.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.436;TN911.7

【参考文献】