题目：多物理场测量和贝叶斯方法报告

时间：2022年10月30日 14:00-16:00

腾讯会议会议号：333-997-014

邀请人：余亮 博士（振动、冲击、噪声研究所）

报告题目：Bayesian Inference for Inverse Problems and multi-modal image fusion

报告人：阿里 嘉法理 教授

报告人简介

阿里 嘉法理，教授。曾任法国国家科学研究中心研究主任，法国巴黎第十一大学终身教授，智能探测领域资深专家。他提出的高斯-波特图像分割算法、快速贝叶斯变分法、超参数贝叶斯推断方法等，受到国际学术和工业界高度认可。相关研究成果在无损探测、机械故障诊断、医学图像识别和工业大数据分析领域等广泛应用，被欧洲空中客车、法国泰勒斯、达索飞机、法国核电等直接采用。发表论文300余篇、2部专著、12本教材、培养博士生21位、硕士生31位。现为上专股份首席科学家、总工程师，首位浙江省引进的产业类国际顶尖人才。

报告摘要

Classical deterministic methods for inverse problems are mainly based on regularization theory. The Bayesian approach gives more flexibility in choosing these terms via the likelihood and the prior probability distributions. In this talk, particular examples of classical inverse problems, such as image denoising, more difficult image restoration and Computed Tomography image reconstruction will be considered. Then, a multimodal image fusion encountered in Non Destructive Testing is considered and different forward modeling and Bayesian solutions are presented.

报告题目：Bayesian inference in acoustic source localization and temperature reconstruction

报告人：初宁博士

报告人简介

初宁博士，获得法国巴黎十一大学信息与通信系统科学博士学位，曾任瑞士洛桑联邦理工合作研究员，现任上专股份首席研究员，副总工。长期从事贝叶斯推断方法，多传感器信息融合，在国际上率先研发工业芯肺系统，推动传统通风装备向绿色智能化转型，荣获国家四部委智能制造优秀案例。近5年发表高水平SCI论文32篇，以第一完成人授权发明专利28项，国际专利2项，软著6项，荣获中国职责安康协会科技进步二等奖(排2)。

报告摘要

We present Bayesian inference in acoustic source localization and temperature reconstruction. Noise source localization can effectively serve condition monitoring and machine noise reduction. The algorithms under Bayesian framework with Student-t priors are deployed to solve acoustic inverse problem in non-synchronous measurement for lower frequency acoustic localization. Infrared imaging is widely used to detect heat source thanks to its non-contact and large-scale thermography. However, measured temperature is always less than the actual value. We propose a novel method of temperature reconstruction based on Bayesian inference.