基于支持向量机的地震事件类型自动识别及应用

(福建省地震局,福州 350003)

事件类型;特征组合;自动识别模块;支持向量机

Automatic Identification of Earthquake Event Types Based on Support Vector Machine and Its Application
CAI Xinghui,LIAO Shirong,ZHANG Yanming,CHENG Huifang,LIN Binhua

(Fujian Earthquake Agency,Fuzhou 350003,China)

Event type;Feature combination; Automatic recognition module;Support Vector Machine

DOI: 10.13512/j.hndz.2021.02.04

备注

天然地震与非天然地震自动识别是地震自动编目系统的重要功能之一,是监测数据产出智能化的基础应用。从福建天然地震和人工爆破事件中,提取小波分析特征、P/S震相振幅比、波形能量分布特征,对以上特征组合联合支持向量机进行大批量数据测试分析,研究得出识别效果较好的事件类型判别算法,最优测试识别率为94.5%;采用最优算法研发基于支持向量机事件类型自动识别模块,将研制的自动识别软件模块应用于福建台网的日常地震编目工作,对2019年8月1日至2020年2月29日共计1531个日常触发事件进行准实时分类,自动识别软件模块分类的正确率为93.9%。另外,采用AMQ消息中间件为信息中介,从redis波形共享内存中获取事件波形,实现自动识别模块与自动编目系统对接。

Automatic identification of natural and non-natural earthquakes is one of the important functions of earthquake automatic cataloging system,and it is the basic application of intelligent monitoring data output. The wavelet analysis features,P/S phase amplitude ratio,waveform energy distribution features and spectrum features were extracted from the natural earthquake and artificial blasting events in Fujian province to carry out large-scale data test and analysis on the combination of the above features and the Support Vector Machine,the algorithm of event type identification is proved to effective,and the optimal test recognition rate is 94.5%. An automatic event recognition module based on Support Vector Machine is developed by using the optimal algorithm,the automatic identification software module was applied to the daily earthquake cataloging work of Fujian Seismic Network,and a total of 1531 daily triggered events were quasi real time classified from August 1st,2019 to February 29th,2020,the correct rate of automatic recognition software module classification was 93.9%. Using Amq message middleware as the information medium,the event waveform is obtained in Redis waveform shared memory,and the automatic recognition module is connected with automatic cataloging system.

·