基于深度学习的光纤网络链路缺陷数据检测及研究

Data detection and research of optical network link defects based on deep learning

为保证光纤网络的正常通信,有必要进行缺陷数据检测。然而,由于数据的不平衡性,导致检测质量受到影响。针对上述问题,研究一种基于深度学习的光纤网络链路缺陷数据检测方法。该方法先进行光纤网络链路缺陷数据预处理,包括去噪处理、插值、均衡处理、尺度处理、时长处理等,提高数据质量,然后利用基于规则迭代的方法进行时间序列特征提取,包括规则产生、规则迭代以及特征提取三步骤,最后以深度学习中的深度置信网络为基础,结合支持向量描述方法,构建一个缺陷判断模型,实现光纤网络链路缺陷数据检测。结果表明:所提方法应用下,ACC值达到90%以上,DR值达到80%以上,FAR值仅为2%左右,证明所提方法检测质量较高,解决了因数据的不平衡性引发的检测质量问题。

In order to ensure the normal communication of optical fiber network,it is necessary to detect the defect data.However,due to the imbalance of data,the detection quality is affected.In order to solve the above problems,this paper studies the data detection method of optical network link defects based on deep learning.Firstly,the data preprocessing of optical fiber network link defects is carried out,including denoising,interpolation,equalization,scale processing,time length processing,etc.to improve the data quality,then the rule-based iterative method is used to extract time series features,including rule generation,rule iteration and feature extraction.Finally,it is based on the deep confidence network in deep learning Combined with the support vector description method,a defect judgment model is constructed to realize the defect data detection of optical fiber network link.The results show that:under the application of the proposed method,ACC more than 90%,DR more than 80%,and the FAR value is only about 2%.It proves that the proposed method has high detection quality and solves the detection quality problems caused by the imbalance of data.