News Source & Photographer: School of Optics and Photonics

Editor: Sheng Yun

Reviewer: Dong Liquan

Translator: Wen Lin

Recently, the team of Xu Tingfa, the professor of BIT, in cooperation with The National Astronomical Observatories of the Chinese Academy of Sciences, have invented the first automatic patrol and detection system for FAST’s reflective surface through deep combination of unmanned system and AI as well as innovative breakthrough in core technology in underlying platform and top-level algorithm, successfully achieving the automation and precision in the process of patrol and detection of reflective surface. The related research results entitled “Automated optical inspection of FAST’s reflector surface using drones and computer vision” were published in Light: Advanced Manufacturing, a a high starting point Chinese journal of the Excellence in Science and Technology Action Plan. As a derivative of Light:Science&Applications, a leading journal belongs to Nature, Light: Advanced Manufacturing focuses on the publication of the world’s latest achievements in the realm of advanced manufacturing and showing interdisciplinary cutting-edge technologies and development trends. Li Jia’nan, the associate research fellow of BIT, is the first author of this paper and shares the position of corresponding authors with BIT Professor Xu Tingfa.

Five-hundred-meter Aperture Spherical radio Telescope (FAST), also crowned as “China's Eye of Heaven”, is the world's largest single-aperture and the most sensitive radio telescope with China's independent intellectual property rights. The main body of FAST is a five-hundred-meter spherical reflective surface composed of 4,450 units of single reflective surfaces (Fig.1a), covering a total area of approximate 250,000 square meters. The gigantic reflective surface of FAST shows unprecedented super sensitivity, but therefore being vulnerable to the damages caused by natural falling objects like falling rocks and hails, which result in panel defects such as dents and holes (Fig.1c), devastating the functional reliability of FAST. So precisely detecting and repairing the panels’ detects in time is of great significance to FAST’s stable functionality. Traditional patrol and detection process mainly depends on periodical manual operations of every reflective surface unit. Influenced by the panels’ weight-bearing capacity, weather and risks of working at height, manual inspection is flawed with high error rate, poor traceability and low efficiency.

Targeted at difficulties above, BIT research team of Professor Xu Tingfa designed the first automatic patrol and detection system for FAST’s reflective surface and successfully guaranteed automated and precisive inspection process, reducing the identification error rate and risks, greatly improving the process efficiency to protect the long-term and stable performance of FAST.

Fig. 1 Schematic diagram of FAST’s structure, intelligent patrol and detection process, results.

Based on multi-rotor UAV (unmanned aerial vehicle) platform and navigation scheme with high positioning accuracy, this inspection system can comprehensively utilize technology of UAV, visual positioning, deep learning and then guide UAVs to fly along scheduled routes and return relevant data. Based on remote digital management platform, the system can identify data gathered in flights with AI and generate inspection reports. In this way, it helps maintenance personnel to find panel defects more quickly and repair them in time. To solve the problem of missed and false detection of subtle and confusing panel defects, the research team proposes an intelligent detection method for subtle defects in complex backgrounds by introducing a pixel-by-pixel cross-layer attention mechanism into the depth detection network, which improves detection accuracy.

The system can carry out FAST reflective panel inspection autonomously and efficiently and achieve centimeter-level precise identification and positioning (Fig.1d), thus having a profound significance for the long-term and stable operation of FAST.

Article details: Li J, Jiang S, Song L, et al. Automated optical inspection of FAST’s reflector surface using drones and computer vision[J]. Light: Advanced Manufacturing, 2023, 4(1): 1-11.

Article link:

https://www.light-am.com/article/doi/10.37188/lam.2023.001

About the authors:

Dr. Li Jia’nan, is a pre-hired assistant professor (special associate researcher) of School of Optics and Photonics, BIT and a postdoctor of National University of Singapore. He has been devoted to the research of photoelectric imaging detection and recognition for a long time. As the first author, he has published 10 papers in top meeting and top journal such as IEEE TPAMI, CVPR, including a highly-cited ESI paper. As a corresponding author, he has published about 20 papers on IEEE TMI, NeurIPS, etc. The highest number of single citation of his paper has reached about 700, and his papers have totally been cited 3,000 times in Google Scholar.

Dr. Xu Tingfa, BIT professor and doctoral supervisor, is the responsible professor of “Optical Engineering” discipline (a national first-level key discipline) ,the deputy director of optoelectronic imaging technology and system laboratory (a key laboratory of the Ministry of Education), and the director of intelligence and big data technology laboratory of Chongqing Innovation Center of BIT. In recent years, he has been leading his research team to further study on photoelectronic imaging detection and identification, computational imaging, AI, etc. He has hosted and undertaken about 40 key research projects of The National Natural Science Foundation of China for the development and manufacturing of research instruments. He has published about 170 academic papers in several international and domestic journals, among which about 90 papers were included in SCI or EI. As the first inventor, he has applied for 45 national invention patents, in which 15 have been authorized and publicized. He has won 3 prizes including provincial and ministerial-level second prize for science and technology progress. He instructed postgraduates to accomplish excellent doctoral dissertations of Chinese Society of Image and Graphics so that the two received Wang Daheng College Student Optics Award and gained the title of Top 100 of National Optics and Optical Engineering Doctoral Academic League, Chongqing Yingcai Innovation and Entrepreneurship Demonstration Team and so on.