Deep Learning for Polarization Optical System Automated Design

• DOI: 10.3390/photonics11020164
• Publication Date: 2024-02-08T08:36:17Z
• Authors: Haodong Shi; Ruihan Fan; Chunfeng He; Jiayu Wang; Shuai Yang; Miao Xu; Hongyu Sun; Yingchao Li; Qiang Fu
• Keywords: Optical engineering

Aiming at the problem that traditional design methods make it difficult to control polarization aberration distribution of optical systems quickly and accurately, this study proposes an automatic optimization method for based on deep learning. The unsupervised training model ray tracing polarized was constructed by learning reference lens structural feature data from library, generalization ability neural network improved achieve system. results show system structure optimized is close in full field view spectrum can be designed different focal length requirements. success rate 1 million groups initial structures better than 96.403%, effect effectively controlled. proposed approach provides a new solution future complex also effective way improve accuracy special such as systems.