双曲Radon域预测反褶积和反馈循环的多次波压制综合方法

A synthesis method of multiple elimination with predictive deconvolution and iterative feedback in hyperbolic Radon domain

针对预测反褶积未能有效压制中长周期或速度逆转的多次波,以及反馈循环法在未知声源子波和地层结构情况下多次波压制不显著等问题,提出了双曲Radon域预测反褶积和反馈循环的多次波压制综合方法,有效突破了两种方法的局限性.首先分别从预测反褶积和反馈循环两种途径进行处理预测出多次波成分,然后进行双曲Radon变换,由于预测反褶积预测误差中仍然存在长周期多次波,通过设计自适应滤波器将有效一次波滤掉,进而获得双曲Radon域多次波能量模型;再通过双曲Radon反变换,从原始浅地层剖面记录中减去多次波模型数据,获得多次波压制后的高可靠性浅地层剖面数据.实验证明,综合法压制多次波效果显著,峰值信噪比和图像结构相似度相对单独使用预测反褶积或反馈循环法分别高出约5%和0.12.该方法具有较好的可行性,在海底浅表层多次波压制方面具有借鉴意义.

The predictive deconvolution could not effectively eliminate the long period or velocity reversal multiple,and the iterative feedback method also could not obtain significant result after eliminating the multiple in case of not knowing the truth of sound source wavelet and the structure of submarine sediment.To solve this problem,a synthesis method of multiple elimination with predictive deconvolution and iterative feedback in hyperbolic Radon domain was proposed,which effectively breaks through the limitations of the two methods.The multiple was distinguished from raw data by the methods of predictive deconvolution and iterative feed-back respectively,then the multiple was handled with hyperbolic Radon transform.Since the long period multiple still exists in the prediction error from predictive deconvolution,the multiple energy model in hyperbolic Radon domain was obtained by means of the self-adapting filter to eliminate the effective primary,and then through the inverse hyperbolic Radon transform,high reliability sub-bottom profile data was acquired via subtracting the multiple energy model data from the raw data.The experiments show that significant result for eliminating multiple can be obtained by the synthesis method,and peak signal to noise ratio and image structure similarity were about 5% and 0.12 higher than predictive deconvolution or iterative feedback used alone.The synthesis method is more feasible and provides the valuable experience of eliminating the multiple in sub-bottom profile data.