光纤振动信号识别的线性尺度不一致校正方法
|
摘要
针对光纤振动信号存在线性尺度不一致的问题,本文提出了一种校正方法.首先,对数据进行预处理并截取等长的单个振动信号形成数据库;然后,将存在尺度不一致的信号进行梅林变换,使信号还原到同一尺度,得到不变特征的提取;最后,对数据进行傅里叶变换,获得频谱特征后,将校正前后的数据送入BP神经网络中进行分类.实验采用镐刨、挖地、小跑3种信号进行验证,校正后数据识别率分别提高了4%、10%和12%.实验结果表明该方法能够有效解决光纤振动信号尺度不一致问题,并与未校正数据相比,校正后数据进一步提高了信号识别率.
For the problem of linear inconsistency in the optical fiber vibration signal,this paper puts forward a correction method.First,the data interception is performed in the vibration signals with equal length to form a database after preprocessing.Then,the signal of the scale inconsistency is transformed with Mellin transform to restore the signal to the same scale so as to get the invariant features extraction.Finally,the Fourier transform is adopted to obtain the spectral characteristics and the data before and after correction are further sent into BP neural network for classification.The experiment uses three kinds of signals,such as picking,digging and running,The data recognition rate after correction is increased by 4%,10% and 12% respectively.The experimental results show that the method can effectively solve the problem of scale inconsistency in fiber vibration signal,and the signal recognition rate is improved with the corrected data compared with the uncorrected data.
For the problem of linear inconsistency in the optical fiber vibration signal,this paper puts forward a correction method.First,the data interception is performed in the vibration signals with equal length to form a database after preprocessing.Then,the signal of the scale inconsistency is transformed with Mellin transform to restore the signal to the same scale so as to get the invariant features extraction.Finally,the Fourier transform is adopted to obtain the spectral characteristics and the data before and after correction are further sent into BP neural network for classification.The experiment uses three kinds of signals,such as picking,digging and running,The data recognition rate after correction is increased by 4%,10% and 12% respectively.The experimental results show that the method can effectively solve the problem of scale inconsistency in fiber vibration signal,and the signal recognition rate is improved with the corrected data compared with the uncorrected data.
引文
[1]Qiu Zezheng,Zheng Tong,Qu Hongquan,et al.A new method based on CFAR and DE for OFPS[J].Photonic Sensors,2016,6(3):261-267
[2]Liang Wei,Lu Linlin,Zhang Laibin.Coupling relations and early-warning for“equipment chain”in long-distance pipeline[J].Mechanical Systems&Signal Processing,2013,41(1-2):335-347
[3]朱程辉,瞿永中,王建平.基于时频特征的光纤周界振动信号识别[J].光电工程,2014,41(1):16-22
[4]杨正理,孙书芳.基于小波能量熵的光纤周界安防系统信号识别[J].光电子·激光,2016(12):1328-1333
[5]曲洪权,陈雨佳,邢志强.基于小波变换的光纤振动信号特征提取与识别[J].北方工业大学学报,2015,27(3):20-24
[6]王照勇,潘政清,叶青,等.用于光纤围栏入侵告警的频谱分析快速模式识别[J].中国激光,2015(4):159-164
[7]孙茜,封皓,曾周末,等.基于图像处理的光纤预警系统模式识别[J].光学精密工程,2015,23(2):334-341
[8]Xiao B,Ma J F,Cui J.Combined blur,translation,scale and rotation invariant image recognition by Radon and pseudo-Fourier-Mellin transforms[J].Pattern Recognition,2012,45(1):314-321
[9]Cheng J,Gao G,Ding W,et al.An improved scheme for parameter estimation of G°distribution model in high-resolution SAR images[J].Progress in Electromagnetics Research,2013,134:23-46
[2]Liang Wei,Lu Linlin,Zhang Laibin.Coupling relations and early-warning for“equipment chain”in long-distance pipeline[J].Mechanical Systems&Signal Processing,2013,41(1-2):335-347
[3]朱程辉,瞿永中,王建平.基于时频特征的光纤周界振动信号识别[J].光电工程,2014,41(1):16-22
[4]杨正理,孙书芳.基于小波能量熵的光纤周界安防系统信号识别[J].光电子·激光,2016(12):1328-1333
[5]曲洪权,陈雨佳,邢志强.基于小波变换的光纤振动信号特征提取与识别[J].北方工业大学学报,2015,27(3):20-24
[6]王照勇,潘政清,叶青,等.用于光纤围栏入侵告警的频谱分析快速模式识别[J].中国激光,2015(4):159-164
[7]孙茜,封皓,曾周末,等.基于图像处理的光纤预警系统模式识别[J].光学精密工程,2015,23(2):334-341
[8]Xiao B,Ma J F,Cui J.Combined blur,translation,scale and rotation invariant image recognition by Radon and pseudo-Fourier-Mellin transforms[J].Pattern Recognition,2012,45(1):314-321
[9]Cheng J,Gao G,Ding W,et al.An improved scheme for parameter estimation of G°distribution model in high-resolution SAR images[J].Progress in Electromagnetics Research,2013,134:23-46