太阳射电频谱图像自动实时检测研究

Solar radio observation is one of the important methods to study the solar eruption. A large number of solar radio bursts and fine structures in the dynamic spectrum come from solar eruption, and their characteristic parameters reflect the important physical parameters of the solar active region. With the development of solar radio spectrometer, a large number of observational data has been obtained and the artificial detection is difficult to reach the research needs. Therefore, it is very important to extract the important characteristic parameters from a number of dynamic spectra rapidly and efficiently..This project uses intelligent visual surveillance technology to detect and recognize solar radio spectrum automatically. Firstly, the regions of the spectrum’s pixel change are detected, and these regions show that the corresponding solar radio bursts are found. Secondly, the feature parameters of solar radio bursts are extracted, including frequency drift rate, burst bandwidth, burst duration, harmonic ratio, burst intensity, etc. Thirdly, according to the feature parameter extracted, solar radio bursts and fine structures are automatically classified using artificial neural network and machine learning techniques. Finally, the index files of solar radio bursts are established in order to provide convenience for the subsequent research work..This technology can be used to automatically process and recognize the observational data from Mingantu Ultrawide Spectral Radio Heliograph(MUSER), and can also provide a new idea for other astronomical image processing.

太阳射电频谱观测是研究太阳爆发活动的重要手段。太阳爆发活动中产生的大量射电频谱精细结构，其频谱特征参数直接或间接地反映了太阳活动区的重要物理参数。随着射电频谱仪的发展，观测数据呈海量趋势，人工检测已难以满足研究工作的需求。因此，如何高效迅速地从这些海量信息中提取关键的太阳射电爆发频谱特征参数，显得十分重要。.本项目将智能视频监控的思想用于对太阳射电频谱图像的自动检测与识别。首先检测频谱图像像素变化的区域，这个区域对应某个太阳射电爆发；然后提取出需要的爆发特征参数值（包括漂移速度、爆发带宽、持续时间、谐波比例和强度等）；再利用人工神经网络和机器学习技术，对检测到的射电爆发的类型及其精细结构进行自动分类；最后建立太阳射电爆发的索引文件，为相关研究工作提供便利。.该项技术今后可以用于明安图射电频谱日像仪观测资料的频谱自动处理及识别，也可以为其它天文图像处理方法提供新的思路。

本项目的研究通过计算机、采用数字图像处理的技术，对观测得到的太阳射电频谱图像进行自动、实时地检测，自动判断有无太阳射电爆发，对有爆发的频谱图像自动提取爆发的特征参数值，对后续太阳物理理论研究工作提供很大便利和帮助。.项目执行期间，项目组提出了一种太阳射电频谱图像的去噪方法，提出了三种太阳射电爆发自动检测和相关参数的自动提取方法，提出了对太阳射电爆发进行预警，降低太阳射电爆发对卫星和导航的影响的方法。这些方法进行了实验，取得了良好的效果，可以为其它天文图像处理方法提供新的思路。.项目负责人到中科院国家天文台明安图观测基地（中国科学院太阳活动重点实验室）进修访学一年，正在参与明安图射电频谱日像仪观测资料的频谱自动处理及识别。