경희대학교 BK21

아래와 같이 세미나를 진행하오니
관심있는 분들의 많은 참여를 바랍니다.
-------------------- 아 래 --------------------
● 날짜 : 2015년 11월 17일 (화요일)
● 시간 : 16시 30분
● 장소 : 천문대 영상실
● 강사 : Dr. N. Nishizuka (NICT)

Title :

Solar Flare Prediction Technique Based on Image Processing of Real-time Solar Magnetogram

 
Abstract :

Currently used solar flare prediction techniques are mostly based on the statistics of sunspot features by white light observations. It is known that big and complex sunspots are more likely to produce large flares with high probability. Recent space solar ations,
like Hinode by JAXA/NAOJ and SDO by NASA, have enabled the steady near real-time observations of vector magnetogram in the
photosphere, i.e. three components of magnetic field in the sunspots. The data of vector magnetogram shows us the direct diagnostics of the amount of stored magnetic energy around the sunspot and the appearance of small scale magnetic field elements triggering flares below the photosphere. Therefore it may be possible to predict solar flares with larger probability and higher accuracy in the way of automatic image processing of magnetogram data in the near real time. Now we are developing a new method to predict solar by analyzing the magnetogram images taken by HMI telescope on board SDO with the image processing techniques. At first, we detect sunspot regions from a magnetogram image, where the flare occurrence is
highly expected. Secondly, we calculate the characteristic values of the region: the physical characteristic values like the sunspot area, the maximum and minimum magnetic field strength and the share angle of the magnetic field lines along the polarity inversion line, and the image characteristic values of the sunspots like colors and shapes. We have developed a system to automatically detect the sunspot regions and calculate the characteristic values, and made a database of these values for 5 years data during June 2010- July 2015. We statistically investigated the tendency of the magnetic field feature of the sunspot just before large flares, and we tried to find which parameters are important to predict flare occurrence and what are threshold values. Finally, we have tried to predict solar flares with machine-learning of these databases of characteristic values and flare occurrences. In this talk, I would like to introduce our developing flare prediction models and techniques.