Solar Radio Observation Data Processing And Recognition Technology

With the increasing resolution of astronomical radio signal observations and the continuous construction of more observatories,the fast and effective processing of massive observation data has become one of the bottlenecks that need to be broken in the field of radio astronomy research.As the main presentation form of solar radio observation data,spectrograms contain important information such as the type,intensity and duration of solar outbursts.Due to the influence of various external military and civilian communication signals,space electromagnetic noise and observation equipment’s own noise,there is a large amount of noise interference in ground-based solar radio observation signals,which greatly affects the fine observation of solar activity.Based on practical research projects,this project analyzes the current status of domestic and international research on spectrogram noise reduction and solar outburst detection,and carries out research and implementation of solar radio spectrogram noise reduction and outburst detection algorithms using deep learning technology based on the massive data of solar radio observation signals,in response to the problem that traditional image processing methods cannot effectively process solar radio spectrograms.The main research contents and contributions include:(1)Aiming at the problem that the traditional image denoising method cannot effectively reduce the noise in the solar radio spectrum,a noise reduction method for the solar radio spectrum based on the CBDNet noise reduction network model is proposed.Collect and organize the spectrograms from three sources: Yunnan Astronomical Observatory,Culgoora Observatory and Learmonth Observatory to build a noise reduction data set,train the noise reduction network model and conduct experiments.The results prove that the noise reduction method based on the CBDNet network model has good noise reduction effect and noise reduction efficiency.(2)Aiming at the low efficiency of traditional manual detection of solar radio burst events,a solar radio spectrum burst detection method based on YOLOv5 s target detection network is proposed to improve detection efficiency and reduce labor costs.Organize the spectrogram containing type III outbreak events to create a target detection data set,train the target detection network model and conduct experiments.The results prove that the YOLOv5 s target detection network achieves an average detection accuracy of 96%.Compared with other detection algorithms,the YOLOv5 s model has excellent detection speed,detection accuracy and lightweight model,which meets the needs of solar eruption event detection.(3)Aiming at the requirement of low frequency radio observation,a system architecture combining PC and FPGA is proposed to complete the software design and implementation of solar radio burst real-time monitoring system.The user interaction interface,transmission protocol and function module workflow of solar radio signal acquisition and control software are designed in detail,and the solar radio spectrum denoising and outbreak detection algorithm are integrated with the software,completing the design,implementation and testing of the solar radio outbreak real-time monitoring system.