激光网络光栅坏点数据检测  

Data detection of grating bad point in laser network

作　　者：苗金萍[1] MIAO Jinping(Qinghai Animal Husbandry and Veterinary Vocational Technical College, Qinghai 812100, Chin)

机构地区：[1]青海畜牧兽医职业技术学院

出　　处：《激光杂志》2018年第5期116-119,共4页Laser Journal

摘　　要：由于激光网络数据量的飞速增长，采用传统光栅坏点数据检测方法已经无法满足检测精度的要求，为此，提出基于密度聚类的激光网络光栅坏点数据检测方法。所提方法结合非监督学习挖掘技术，确定光栅坏点所在聚类区间。利用密度聚类对获得的激光聚类区间进行合理划分，根据曼哈顿定理得到数据挖掘距离定义式，引入权重数值，根据熵特征的权重冗余度和相关性得到比重备选集合，确定重要性权重具体数值。在得到的某个聚类区间中，如果出现了特征差异较大且数据挖掘的距离十分突出的情况，那么可以确定其为光栅坏点数据。通过实验结果的分析可以证明，多提方法实用性好，对光栅坏点的检测精度较高。For the rapid growth of laser network data, traditional data detection of grating bad point in laser network can not meet the requirement of accuracy. Therfore, a novel method based on density clustering is proposed. This method combines unsupervised learning mining technology and ensure the clustering section of grating bad point. Using desity clustering to divide the laser clustering section reasonably, obtaining the defination formula of data mining distance according to Manhattan theorem, introducing weighted value, getting alternative proportion sets through the weighted redundant and correlation of entropy features and ensuring certain value of significant weight. In one obtained clustering section, if there is situation that the feature difference is big and the distance of data mining is long, it can be sure as the grating bad data. The experimental analysis indicates this method has good feasibility and accuracy.

关 键 词：激光网络 光栅坏点 数据检测 非监督学习技术 密度聚类 

分 类 号：TN929[电子电信—通信与信息系统]