What is the performance of the telescope based on machine learning?
Visit: Alibaba Cloud “explosive special offers” theme activity-cloud server as low as 0.55 fold 96 yuan/year
The performance monitoring of the telescope refers to the optical performance of the telescope during the observation of the telescope. Point to the performance of tracking performance for evaluation. Once the telescope’s performance monitoring system finds that the telescope’s performance is poor, it can quickly feed back the cause of the telescope’s poor performance to the maintenance personnel, thereby improving the maintenance efficiency of the telescope and obtaining excellent observation data quality.
Traditional telescope monitoring methods use sensors to read and record the values of key parts of the telescope. When the values exceed the set threshold, the sensor will warn the maintenance personnel to stop observation. This method can help the telescope avoid risks and Maintenance is aided, but the performance of the telescope cannot be monitored in real time, maintenance and operation efficiency is low, and it is urgent to improve.
Recently, the research team of Nanjing Institute of Astronomical Optics Technology who participated in the operation and maintenance of LAMOST, based on years of telescope maintenance experience, combined with the wide application of artificial intelligence in various fields, proposed a new telescope performance monitoring method. The method is based on the profound and complex correspondence between the shape of the star image obtained by the telescope terminal instrument and the performance of the telescope and the latest cutting-edge machine learning related algorithms. It can make full use of the star image obtained by the telescope for training and testing, and realizes the optical imaging performance of the telescope. High-precision real-time monitoring of astronomy, and a large number of tests and verifications have been carried out on the Guo Shoujing Telescope LAMOST (Large Sky Area Multi-Object Spectroscopic Telescope), China’s first astronomical major national scientific and technological infrastructure Experiments have achieved good practical application effects.
LAMOST has China’s independent intellectual property rights and the original world’s largest wide-field telescope (Wang-Su Active Reflecting Schmidt System) broke the international bottleneck that cannot have both large-field and large-aperture, leading the world’s spectral survey trend. LAMOST has an equivalent clear aperture of 3.6-4.9 meters and a field of view diameter of 5 degrees, which can simultaneously obtain the spectrum of 4000 celestial bodies. It was completed in 2009 and passed the national inspection and acceptance. Since the pilot sky survey and official sky survey began in 2011, it has been in stable operation for 9 years. Through the efforts of the operation and maintenance team composed of the National Astronomical Observatory, Nanjing Institute of Astronomy, University of Science and Technology of China, Shanghai Astronomical Observatory and other units, the hardware status is stable, and the failure rate from 2019 to 2020 is 0.66%. As of 2020, the release of DR7 data reaches 14.48 million. Published nearly 700 SCI papers.
The star images obtained by the terminal instruments of the ground-based astronomical telescope are inevitably affected by atmospheric disturbances and the telescope system. The star images obtained by long-term exposure under normal telescope performance become a circular Gaussian contour. The poor performance of the telescope will cause the obtained image spot shape to deviate from the normal standard circle. The poor performance of the telescope caused by different reasons produces different image spot shapes, so the image spot shape obtained by the terminal instrument can be used to monitor the telescope performance in real time.
Figure 2 Guide Various star shape statistics obtained by star cameras (the first line is the normal observation star image, most of the astrological images)
Machine learning related algorithms have shown good shape recognition and shape recognition in many fields. The classification capability can be used to distinguish the shape of the image spot obtained by the telescope. Combined with machine learning related algorithms, the performance of the telescope can be monitored through the following steps. First, classify the images obtained by the telescope to filter out the pictures containing bright stars; then cut the pictures to obtain small pictures containing only a single complete star image , Use machine learning algorithms to recognize the shape of the star image, and give the reason for the poor performance based on the recognized shape; finally use probability statistics and combine the star image classification results of multiple cameras to give the final reason.
Figure 3 Telescope Monitoring method implementation flow chart
The research team used this method to monitor the performance of LAMOST, and realized failures and problems such as focal plane defocusing, guide star system, stitching mirror sub-mirror offset, active optical performance, etc. Real-time monitoring has an accuracy rate of 96.7%, which improves the operation and maintenance efficiency of LAMOST and helps improve the data quality of the telescope. Because this method does not require installation of additional equipment, it can be easily extended to other telescopes. Relevant research results were published in the Monthly Bulletin of the Royal Astronomical Society (MNRAS).
Figure 4 in Performing performance testing on LAMOST telescope, real-time monitoring of focal plane defocusing, guide star system, splicing mirror sub-mirror offset, and active optical performance was achieved, with an accuracy rate of 96.7%.