Page 161 - 《应用声学》2022年第1期
P. 161
第 41 卷 第 1 期 谢辉武等: 利用 U-Net 网络增强骨密度全波形反演 157
[4] Whitmore N D. Iterative depth migration by backward
3 结论 time propagation[M]//SEG Technical Program Expanded
Abstracts 1983. Society of Exploration Geophysicists,
与骨密度检测的“金标准”双能 X 射线比较来 1983: 382–385.
说,声学法仪器便宜,无辐射,有可能小型化后方便 [5] Clapp R G. Reverse time migration with random bound-
家庭进行提前检查。而定量声学检测法的准确度还 aries[G]//SEG Technical Program Expanded Abstracts
2009. Society of Exploration Geophysicists, 2009:
需要实践的检验。如果使用本文的方法可以对全波
2809–2813.
形骨密度测量的准确度有一定的提高,还可以减小 [6] Ronneberger O, Fischer P, Brox T. U-Net: convolutional
反演对初值的依赖。不过现有的骨密度检测的标准 networks for biomedical image segmentation[J]. Computer
是骨矿物质密度,使用的是骨头对光子的吸收能力 Vision and Pattern Recognition, 2015, arXiv: 1505.04597.
[7] Yang F, Ma J. Deep-learning inversion: a next generation
相对值。使用声学检测需要在未来进一步研究声学
seismic velocity-model building method[J]. Geophysics,
的相对标准或它与光学法的结果对应关系。 2019: 1–133.
[8] Zhu W, Beroza G C. PhaseNet: a deep-neural-network-
based seismic arrival-time picking method[J]. Geophysical
参 考 文 献 Journal International, 2019, 216(1): 261–273.
[9] Aster R C, Borchers B, Thurber C H. Parameter estima-
[1] Warner M, Guasch L. Adaptive waveform inversion: the- tion and inverse problems[M]. Second Edition. Salt Lake
ory[J]. Geophysics, 2016, 81(6): 429–445. City: Academic Press, 2013.
[2] Yang Y, Engquist B, Sun J, et al. Application of op- [10] 杨积忠, 刘玉柱, 董良国. 基于 Born 敏感核函数的速度、
timal transport and the quadratic wasserstein metric to 密度双参数全波形反演 [J]. 地球物理学报, 2016, 59(3):
full-waveform inversion[J]. Geophysics, 2016: 1–103. 1082–1094.
[3] Bernard S, Monteiller V, Komatitsch D, et al. Ultrasonic Yang Jizhong, Liu Yuzhu, Dong Liangguo. Multi-
computed tomography based on full-waveform inversion parameter full waveform inversion for velocity and den-
for bone quantitative imaging[J]. Physics in Medicine and sity based on Born sensitivity kernels[J]. Chinese Journal
Biology, 2017, 62(17): 7011–7035. of Geophysics, 2016, 59(3): 1082–1094.