报告题目： Adaptive multiscale full waveform inverison using Laplacian pyramid-based misfit function

报 告 人：陈勇 教授  哈尔滨工业大学

报告时间：2025年11月25日 16：30—17：30

报告地点：伍卓群楼研讨室六

校内联系人： 吕俊良 [email protected]

报告摘要：Full waveform inverison (FWI) typically employs a distance metric, such as the ℓ2 or ℓ1 norm, to directly quantify the misfit between observed and synthetic data. However, these approaches are known to suffer from cycle skipping, especially when the initial model lacks low-wavenumber information. Multiscale inversion schemes are an effective means of mitigating this problem, but they require careful frequency selection, particularly when the data are dominated by reflection events. An adaptive multiscale strategy based on the Laplacian pyramid decomposition is introduced, in which both observed and synthetic data are separated into hierarchical low- and high frequency bands. Each level of the Laplacian pyramid is defined by applying Gaussian smoothing to the data, followed by downsampling and upsampling, and then computing the difference between the Gaussian smoothed data at the current level and the upsampled representation from the next level. The final misfit is obtained by summing the ℓ1-norm of the differences between corresponding levels of the decomposed observed and synthetic data. Using this Laplacian pyramid, high-frequency components are progressively removed as the level increases, so each level approximates a low-pass filtered version of the original data. Consequently, the Laplacian misfit function can be interpreted as a multiscale measure. Applications in both acoustic and elastic settings are conducted to demonstrate the effectiveness and robustness of the proposed method.

报告人简介：陈勇，教授，博士生导师。2007 年博士毕业于哈尔滨工业大学，2007 年 10 月至 2011 年 12 月在中国石油大学（北京）地质资源与地质工程博士后流动站工作。现任哈尔滨工业大学脱衣舞 副院长，兼任中国工业与应用数学学会理事、黑龙江省工业与应用数学学会秘书长。科研领域：主要研究方向为非线性反问题理论、算法及其在工程领域的应用，偏微分方程与可积系统理论，主持及参与国家级项目十余项，发表 SCI 检索论文五十余篇，获黑龙江省高校科学技术进步一等奖 2 项。教学领域：主持省级教育教学改革研究项目 3 项，主持校级教学改革项目、思政建设项目多项；在《大学数学》等期刊发表教学论文十余篇。