Date of Award

Spring 2011

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer Science

First Advisor

Frank Y. Shih

Second Advisor

Ju Jing

Third Advisor

Haimin Wang

Fourth Advisor

Alexandros V. Gerbessiotis

Fifth Advisor

Cristian Borcea

Abstract

Study of the solar activity is an important part of space weather research. It is facing serious challenges because of large data volume, which requires application of state-of-the-art machine learning and computer vision techniques. This dissertation targets at two essential aspects in space weather research: automatic feature detection and forecasting of eruptive events.

Feature detection includes solar filament detection and solar fibril tracing. A solar filament consists of a mass of gas suspended over the chromosphere by magnetic fields and seen as a dark, ribbon-shaped feature on the bright solar disk in Hα (Hydrogen-alpha) full-disk solar images. In this dissertation, an automatic solar filament detection and characterization method is presented. The investigation illustrates that the statistical distribution of the Laplacian filter responses of a solar disk contains a special signature which can be used to identify the best threshold value for solar filament segmentation. Experimental results show that this property holds across different solar images obtained by different solar observatories. Evaluation of the proposed method shows that the accuracy rate for filament detection is more than 95% as measured by filament number and more than 99% as measured by filament area, which indicates that only a small fraction of tiny filaments are missing from the detection results. Comparisons indicate that the proposed method outperforms a previous method. Based on the proposed filament segmentation and characterization method, a filament tracking method is put forward, which is capable of tracking filaments throughout their disk passage. With filament tracking, the variation of filaments can be easily recorded.

Solar fibrils are tiny dark threads of masses in Hα images. It is generally believed that fibrils are magnetic field-aligned, primarily due to the reason that the high electrical conductivity of the solar atmosphere freezes the ionized mass in magnetic field lines and prevents them from diffusing across the lines. In this dissertation, a method that automatically segments and models fibrils from Hα images is proposed. Experimental results show that the proposed method is very successful to derive traces of most fibrils. This is critical for determining the non-potentiality of active regions.

Solar flares are generated by the sudden and intense release of energy stored in solar magnetic fields, which can have a significant impact on the near earth space environment (so called space weather). In this dissertation, an automated solar flare forecasting method is presented. The proposed method utilizes logistic regression and SVM (support vector machine) to forecast the occurrences of solar flares based on photospheric magnetic features. Logistic regression is used to derive the probabilities of solar flares occurrence, which are then fed to SVM for determining whether a flare will occur. Comparisons with existing methods show that there is an improvement in the accuracy of X-class solar flare forecasting. It is also found that when sunspot-group classification is combined with photospheric magnetic parameters, the performance of flare forecasting can be further lifted.

Share

COinS