Deep learning enhanced terahertz imaging of silkworm eggs development

Hongting Xiong; Jiahua Cai; Weihao Zhang; Jingsheng Hu; Yuexi Deng; Jungang Miao; Zhiyong Tan; Hua Li; Juncheng Cao; Xiaojun Wu

doi:10.1016/j.isci.2021.103316

Deep learning enhanced terahertz imaging of silkworm eggs development

iScience. 2021 Oct 19;24(11):103316. doi: 10.1016/j.isci.2021.103316. eCollection 2021 Nov 19.

Authors

Hongting Xiong¹, Jiahua Cai¹, Weihao Zhang², Jingsheng Hu³, Yuexi Deng³, Jungang Miao¹, Zhiyong Tan^{4

5}, Hua Li⁴, Juncheng Cao^{4

5}, Xiaojun Wu^{1

2

6}

Affiliations

¹ School of Electronic and Information Engineering, Beihang University, Beijing 100191, China.
² School of Cyber Science and Technology, Beihang University, Beijing 100191, China.
³ College of Engineering, Peking University, Beijing 100191, China.
⁴ Key Laboratory of Terahertz Solid State Technology, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050 China.
⁵ Center of Materials Science and Optoelectronics Engineering, University of Chinese Academy of Sciences, Beijing 100049, China.
⁶ Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan 430074 China.

Abstract

Terahertz (THz) technology lays the foundation for next-generation high-speed wireless communication, nondestructive testing, food safety inspecting, and medical applications. When THz technology is integrated by artificial intelligence (AI), it is confidently expected that THz technology could be accelerated from the laboratory research stage to practical industrial applications. Employing THz video imaging, we can gain more insights into the internal morphology of silkworm egg. Deep learning algorithm combined with THz silkworm egg images, rapid recognition of the silkworm egg development stages is successfully demonstrated, with a recognition accuracy of ∼98.5%. Through the fusion of optical imaging and THz imaging, we further improve the AI recognition accuracy of silkworm egg development stages to ∼99.2%. The proposed THz imaging technology not only features the intrinsic THz imaging advantages, but also possesses AI merits of low time consuming and high recognition accuracy, which can be extended to other application scenarios.

Keywords: Applied physics; Machine learning; Wave imaging.