Improvements on coronal hole detection in SDO/AIA images using supervised classification

Abstract

We demonstrate the use of machine learning algorithms in combination with segmentation techniques in order to distinguish coronal holes and filaments in SDO/AIA EUV images of the Sun. Based on two coronal hole detection techniques (intensity-based thresholding, SPoCA), we prepared datasets of manually labeled coronal hole and filament channel regions present on the Sun during the time range 2011-2013. By mapping the extracted regions from EUV observations onto HMI line-of-sight magnetograms we also include their magnetic characteristics. We computed shape measures from the segmented binary maps as well as first order and second order texture statistics from the segmented regions in the EUV images and magnetograms. These attributes were used for data mining investigations to identify the most performant rule to differentiate between coronal holes and filament channels. We applied several classifiers, namely Support Vector Machine (SVM), Linear Support Vector Machine, Decision Tree, and Random Forest, and found that all classification rules achieve good results in general, with linear SVM providing the best performances (with a true skill statistic of % 0.90). Additional information from magnetic field data systematically improves the performance across all four classifiers for the SPoCA detection. Since the calculation is inexpensive in computing time, this approach is well suited for applications on real-time data. This study demonstrates how a machine learning approach may help improve upon an unsupervised feature extraction method.

Statistical Analysis and Catalog of Non-polar Coronal Holes Covering the SDO-Era Using CATCH

“…There exist multiple approaches to this topic with one of the most popular using a single wavelength, intensity-based threshold approach on EUV observations. Due to the high contrast and the optimal filter sensitivity, the coronal emission line of 11 times ionized iron (Fe XII: 193/195 Å) is often used to extract CHs (e.g., Krista and Gallagher, 2009;Rotter et al, 2012Rotter et al, , 2015Reiss et al, 2015;Caplan, Downs, and Linker, 2016;Boucheron, Valluri, and McAteer, 2016;Hofmeister et al, 2017;Heinemann et al, 2018a). Other intensity based approaches include multi-thermal emission recognition (Garton, Gallagher, and Murray, 2018) and spatial possibilistic clustering (Verbeeck et al, 2014).…”

Rearrangements of Open Magnetic Flux and Formation of Polar Coronal Holes in Cycle 24

“…Using a square root transformation, we preprocessed the maps to allow adequate extraction of low-intensity features. Then, we applied the watershed segmentation, the standard k -means clustering, and the fuzzy C -means algorithm to detect CHs (Reiss et al, 2015;Caplan, Downs, and Linker, 2016). The clustering algorithm estimates a membership function that quantifies pixel intensities to separate bright and dark features in a gray-scale image.…”

Supervised classification of solar features using prior information

“…Distinguishing between filament channels and coronal holes would require to include morphological or geometrical properties such as the ones defined in Reiss et al (2015) or magnetogram information (Scholl & Habbal 2008;Lowder et al 2014).…”