基于SVM的φ-OTDR光纤振动传感系统模式识别方法研究

Investigation on Pattern Method of φ-OTDRFiber Vibration Sensor System Based on SVM

针对φ-OTDR光纤传感器系统对相似扰动目标识别判断的误报问题,通过多特征参量识别目标,则可有效降低误报率。提出一种基于φ-OTDR光纤振动传感系统,提取信号主波能量(MVE)、信号主瓣时延(MVBW)、信号主瓣峰峰值(MDV)、信号主波Q值(MVQV)、信号频谱(SFS)等十一种特征参量作为参考衡量标准,利用二叉树的SVM分类器模型,对10组400次实验进行相似目标识别分类,综合识别率达到90%。该方法通过多角度参量进行目标衡量判别,调高了相似识别的准确性。

In order to solve the problem of false alarm in identifying and judging similar perturbed targets byφ-OTDR optical fiber sensor system,the false alarm rate can be greatly reduced if more feature parameters and multiple angles can be extracted.In this paper,an optical fiber vibration sensing system based onφ-OTDR is proposed.Eleven characteristic parameters,such as main wave energy(MVE),main lobe delay(MVBW),peak value(MDV),main wave Q value(MVQV),signal spectrum(SFS),are extracted as reference criteriaandBinary tree SVM classifier model.Method of recognition,10 groups of 400 experiments were carried out to identify and classify similar targets,and the comprehensive recognition rate reached 90%.This method improves the accuracy of similarity recognition by using multi-angle parameters to measure and discriminate targets.