摘要
文中提出将三帧差分与混合高斯模型相结合的算法来对太阳Hα图像中的特征进行检测。对图像进行标准差归一化后,利用维纳滤波方法对图像进行平滑处理;并利用三帧差分快速得到一副背景图像,更新到混合高斯模型主背景分布中建立稳定的动态背景,从而探测作为前景变化的耀斑、暗条活动;按照太阳活动面积去除掉前景图像中残留的噪声点完成检测操作;根据检测结果计算太阳活动的相关数据。文中分别选用了国家天文台怀柔太阳观测基地的两组数据和太阳全球振荡监测网的一组数据来展示其执行过程并对结果进行评价。实验结果表明,同目前存在的耀斑识别算法相比,文中算法能够自动识别出暗条和耀斑的爆发;能根据耀斑面积演化计算出耀斑级别和耀斑的开始、峰值和结束时刻;有效解决了噪声与光照变化的影响,提高了耀斑检测的准确度。
An algorithm combining the three frame ditterence and the mixed Gauss model was proposed to detect the teatures in the solar Hot image, in this paper. The images were normalized using the standard deviation and the images were smoothed by Wiener filtering. Then, the background image was quickly obtained by using the three frame difference method, then it was updated to the mixed Gaussian mixture models in the main background distribu- tion. Stable Hα images with dynamic background were established. So flares or filaments were detected as solar flares with foreground changes. The residual noise points in the foreground images were removed in accordance with the so- lar activity area and the detection operation was completed. Finally, the related data of solar activity were calculated. This paper was based on the data from Huairou Solar Observing Station （HSOS） and GONG and three sets of data were used to demonstrate its execution and to evaluate the results. The experimental results show that, compared with the existing flare recognition algorithms, it could do three points. The eruption of filaments and flares were automati- cally recognized at the same time. The start, peak and end moments of flares were calculated by the changes in the area and intensity of flares. At the same time flare levels were automatically calculated. The ettect of noise and illu- mination changes had been effectively solved, the accuracy of flare detection was improved.
作者
梁宇峰
白先勇
冯松
LIANG Yufeng;BAI Xianyong;FENG Song(Yunnan Key Laboratory of Computer Technology Applications,Faculty of Infomlation Engineering ＆ Automation,Kunming University of Science ＆ Technology,Kunming 650500,China;Key Laboratoi7 of Solar Activity,Chinese Academy of Sciences,Beijing 100012,China;National Astronomical Observatories,Chinese Academy of Sciences,Beijing 100049,China)
出处
《电子科技》
2018年第12期57-62,67,共7页
Electronic Science and Technology
基金
国家自然科学基金(11463003
11503063)
关键词
太阳活动特征检测
维纳滤波
三帧差分法
混合高斯模型
耀斑等级
暗条识别
solar activity characteristics detection
wiener filtering
three frame ditterence
Gaussian mixture models
flare level
filament recognition
