The Faculty of Information Technology of Macau University of Science and Technology held an academic seminar on June 25, 2021 in Room A201, Block A. We invited Prof. Yunbo Zhang from Zhejiang Sci-Tech University to be the guest speaker of this seminar. He gave a lecture on "Deep Learning Based Optical Surface Impurity Detection".
Abstract:
In modern optical experiments and optical applications based on laser technology, micro-impurities and micro-defects on the surface of optical components are one of the main factors affecting the precision of optical systems, and thus the localization and detection of impurities and defects on optical surfaces is an important issue. We propose a theoretical scheme using deep neural networks to assist in optical impurity detection. Specifically, we simulate that when a probe laser pulse is irradiated to an optical surface with a single tiny impurity, the position information of the impurity carried in the reflected and transmitted signals can be learned and localized by a deep convolutional neural network. A series of generalized data sets are generated by varying the impurity size, refractive index, and other properties, and the performance of the neural network on the generalized data sets is discussed. The results show that the neural network has a high robustness in predicting the impurity locations. Finally, we also compare the learning ability of two different architectural networks, convolutional neural networks and fully connected neural networks.
Biography:
Prof. Yunbo Zhang worked at Shanxi University after receiving his PhD from the Institute of Physics, Chinese Academy of Sciences in 1998, and worked as a Humboldt Fellow at the Kaiserslautern in Germany and the University of Turku in Finland as a postdoctoral researcher. He has been selected as a talent program of the Ministry of Education's New Century Outstanding Talents, Shanxi Province's Academic and Technical Leaders, Shanxi Province's Outstanding Tutor of Postgraduate Education, and Shanxi Province's Sanjin Talent Support Program's Top Backbone Talents. He has published more than 80 SCI papers (including more than 50 in Phys. Rev. series) in important journals at home and abroad. He is mainly engaged in theoretical research on ultracold quantum gases and low-dimensional quantum magnetism, and has achieved important results with international impact in the macroscopic quantum effects of molecular magnets, topological properties of spinor Bose-Einstein condensates, and the critical properties of one-dimensional cold atomic gases with correlations, and has been awarded the Science and Technology Prize of the Ministry of Education and Shanxi Province for four times.
