摘要
基于天然地震和人工爆破的快速识别要求,我们提取了量取方便的初动方向、振幅比、振幅与尾波持续时间比值等5个特征,并判定各单特征的分类能力,其中初动方向和振幅比的识别效果较好,正确识别率分别为91%和92%.为了进一步提高识别效果,本文引入3种模式识别方法和决策方法对多个特征进行综合识别,得到Fisher方法和决策方法均可将正确识别率从92%提高到97%,ICHAM识别效果和最小距离法的综合识别效果次之,可将正确识别率分别提高到95%和94%.结果表明综合识别方法可以不同程度的提高正确识别率,Fisher方法和决策方法在地震和爆破识别中应用效果较好.
Based on the request of rapid identification earthquake and explosion, we extracted 5 criteria： the first polarity.amplitude ratio and amplitude over duration which are all measured conveniently, then we calculated the correct recognition rate of single criterion, among which the first motion polarity and amplitude ratio criterion were better used for discrimination, the correct recognition rate were 91% and 92%. In order to further improve the recognition results, this paper introduced three kinds of pattern recognition methods and decision-making method for comprehensive identification, we obtained that Fisher and decision-making method could both improved the correct recognition ratio from 92% to 97%, ICHAM and the minimum distance method were followed by, could improved the correct recognition ratio to 95% and 94%. Results indicated that comprehensive identification method could improve the correct recognition rate to varying degrees, Fisher and decision-making method were better applied to earthquake and explosion recognition.
出处
《地球物理学进展》
CSCD
北大核心
2013年第5期2433-2443,共11页
Progress in Geophysics
基金
地震行业科研专项基金(200808003)资助
关键词
地震和爆破识别
特征提取
模式识别方法
决策方法
正确识别率
identification of earthquake and explosion
feature extraction, pattern recognition methods, decision-making method, correct recognition rate
