摘要
针对深部找矿地震资料信噪比很低的问题,结合小波包和支持向量机的理论,提出了一种新的去噪算法.该算法首先从地震资料中找出噪声干扰极小的和噪声很强的数据,利用支持向量机进行训练学习,提取噪声和有效信号的样本特征.然后将地震资料实施小波包变换得到更加精细和不同频带的信号,利用支持向量机训练出的样本特征对小波包变换后的数据进行分类提取出有效的信号.最后将有效信号重构还原为地震信号.文章最后通过对实际的金属矿床地震资料进行了去噪,并与其他去噪算法进行比较.结果表明,该算法在理论上是正确的,实际应用效果也是成功的.
A new de-noising algorithm, which is based on the combination of the theories of wavelet package and support vector machine, is raised to solve the problem of low signal-to-noise in seismic data of the metallic ore deposit. In this algorithm, firstly, data with minimal noisy hindrance and data with loud noise would be found out from seismic data, and by using support vector machine training, sample characters about noise and effective signals would be extracted. Then, through implementing the wavelet package transform on the seismic data and making use of the sample characters trained by support vector machine to classify the data after wavelet packet transform effective signals would be extracted. Lastly, these effective signals would be reconstituted to restore the seismic signals. At last, through de-noising practical metallic ore deposit seismic data and comparing with other de-noising algorithrra Finally through a real case analyzed, the result shows that the method is theoretically correct and application effect is successful.
出处
《地球物理学进展》
CSCD
北大核心
2011年第6期2190-2195,共6页
Progress in Geophysics
基金
中国地质调查局地质调查项目(1212010916040)
四川省教育厅自然科学重点项目(08ZA105)资助
关键词
小波包
支持向量机
金属矿床
地震资料
去噪
wavelet packet, support vector machine, metallic ore deposit, seismic data, de-noising
