Automated Recognition And Image Quality Assessment Of Solar Activities Fine Structures

For the improvement of accuracy of the space weather prediction and further study of solar physics,increasing the resolution and objective feature extraction of solar activities fine structures are required.To extract the features of solar activities fine structures accurately and self-consistently,scientists need to make full use of the vast observations from both space-based and ground-based solar telescopes to implement automated detection procedures.In addition,solar adaptive optics system could not compensate atmospheric turbulence completely,so the image quality of consecutive frames may vary.Selecting the images with good quality from consecutive frames could improve the quality of speckle-reconstructed images,and make the object detection and feature extraction easier.In this dissertation,the automated detection methods of the fine structures of sunspots and solar flares,as well as the image quality assessment methods of solar adaptive optics images,are investigated.The main contents of the dissertation are as follows:Firstly,an automated segmentation of sunspot fine structures based on level set method,morphological operations,and gray thresholding is proposed.The previous works are mainly applied to full-disk images,and they could only detect umbra and penumbra instead of the fine structures such as pores and light bridges.Generally speaking,although sunspots are the dark regions within the image,the intensities within each fine structure are not even.The differences between the observations of telescopes all over the world are another trouble of sunspot detection.Fortunately,level set methods could provide closed-form contours as segmentation results and they are robust to intensity inhomogeneity.Level set method could roughly separate solar granulation and sunspot,and then morphological operations and gray thresholding are adopted to detect finer structures.When applied to local high resolution images and full-disk images,the proposed method achieves satisfactory performance.Secondly,an automated solar flare detection method is proposed,and it could be applied to local high resolution and full-disk H? images.Previous works are designed for full-disk images,and the detection methods based on local high resolution images could obtain higher accuracy.The flare detection method adopted in the dissertation use intensity threshold and the area threshold to extract flare region.The solar flare features are extracted,including the start time and the end time,the importance class and the brightness class.Experiments on local high resolution and full-disk H? images demonstrate the effectiveness of this method.Thirdly,an automated detection method of post-flare loops is proposed,and it could separate individual loops from the H? red-wing images.This method combines straight line detection and curve growing,and makes use of the detected flare within the H? core line image.To be specifically,based on the straight line detection result,the deflection orientation of the curve growing process is determined by the slope of straight lines and the position relationship between lines and flares.After the curve growing,fake loops are screened out by the crossing number,deflection orientation and intensity range to increase the accuracy.The effectiveness of this method is demonstrated by the experiments on high resolution H? red-wing images.Finally,an image quality assessment method based on image power spectrum and human visual system is proposed.The widely used contrast methods could only evaluate the quality of solar granulation or sunspot images,while the proposed method could also be applied to large field-of-view images.Experiments demonstrate that the images selected by the proposed method have better quality than contrast methods and other tested methods.Besides,the proposed method has good robustness to image scaling,image blur and noise.In summary,these results and innovations at the field of automated detection of solar activities fine structures are important guidance for improving the accuracy of solar activities feature extraction and space weather prediction,analyzing the evolution of solar activities fine structures,the theoretical investigations of the dynamics of solar activities.The proposed image quality assessment method in this dissertation could help to gain wide field-of-view reconstructed images with better quality,thus brings advantages to the automated recognition of solar activities fine structures.The results are also useful to the automated detection of finer structures of solar activities.