南京理工大学学报(自然科学版)
南京理工大學學報(自然科學版)
남경리공대학학보(자연과학판)
JOURNAL OF NANJING UNIVERSITY OF SCIENCE AND TECHNOLOGY
2014年
6期
757-762
,共6页
胡峰%马春侠%崔毅安%史广顺
鬍峰%馬春俠%崔毅安%史廣順
호봉%마춘협%최의안%사엄순
光纤传感网络%数字信号处理%人工神经网络%通信线路安全
光纖傳感網絡%數字信號處理%人工神經網絡%通信線路安全
광섬전감망락%수자신호처리%인공신경망락%통신선로안전
optical fiber sensing network%digital signal processing%artificial neural network%communication line security
提出了一种基于分布式光纤和人工神经网络判别的通信线路防护系统。该系统利用光纤传感器收集通信线路周围的振动信号,运用数字信号处理的方法对原始信号进行处理,通过神经网络判断是否存在针对通信线路的破坏性行为并判别破坏行为的类型,实现对通信线路的防护。系统在定位阶段,基于Mach-Zehnder干涉原理,运用互相关的方法进行实时定位。在数据处理阶段对信号进行抑噪处理,有利于进一步的定位与事件识别工作。在识别阶段使用支持向量机( Support vector machine,SVM)和反向传播( Back propagation,BP)神经网络方法构建了层次化分类器。实验结果表明:信号定位精度达到100 m,系统对七类破坏行为的识别率达到94.35%。
提齣瞭一種基于分佈式光纖和人工神經網絡判彆的通信線路防護繫統。該繫統利用光纖傳感器收集通信線路週圍的振動信號,運用數字信號處理的方法對原始信號進行處理,通過神經網絡判斷是否存在針對通信線路的破壞性行為併判彆破壞行為的類型,實現對通信線路的防護。繫統在定位階段,基于Mach-Zehnder榦涉原理,運用互相關的方法進行實時定位。在數據處理階段對信號進行抑譟處理,有利于進一步的定位與事件識彆工作。在識彆階段使用支持嚮量機( Support vector machine,SVM)和反嚮傳播( Back propagation,BP)神經網絡方法構建瞭層次化分類器。實驗結果錶明:信號定位精度達到100 m,繫統對七類破壞行為的識彆率達到94.35%。
제출료일충기우분포식광섬화인공신경망락판별적통신선로방호계통。해계통이용광섬전감기수집통신선로주위적진동신호,운용수자신호처리적방법대원시신호진행처리,통과신경망락판단시부존재침대통신선로적파배성행위병판별파배행위적류형,실현대통신선로적방호。계통재정위계단,기우Mach-Zehnder간섭원리,운용호상관적방법진행실시정위。재수거처리계단대신호진행억조처리,유리우진일보적정위여사건식별공작。재식별계단사용지지향량궤( Support vector machine,SVM)화반향전파( Back propagation,BP)신경망락방법구건료층차화분류기。실험결과표명:신호정위정도체도100 m,계통대칠류파배행위적식별솔체도94.35%。
A new communication line protection system has been proposed,which is based on the dis-tributed optical fiber and artificial neural network discrimination. The system uses optical fiber sensors to collect the soil vibration signal around communication line. Raw signals are processed via several kind of digital signal processing methods. A hybrid classification system is applied to identify the existence of destructive behavior. An accurate mutual correlation method is designed based on Mach-Zehnder interference principle to locate the position of vibration signals. Wavelet shrinkage and Hilbert transformation method are applied to filter noise and segment the interest signal section. A two level classifier based on Support Vector Machine ( SVM ) and Back Propagation ( BP ) neural network is designed to identify the type of dangerous behavior. The system has been evaluated under a real application environment. The location deviation is less than 100 m, and the recognition accuracy rate for seven types of dangerous behavior comes to 94 . 35%. The test results prove the efficiency and precision of the system.