A Hough Transform Procedure Applied to Occult-2 Data for Pattern Recognition in SDO/AIA Images

Abstract:

The exploitation of solar data provided by the NASA mission Atmospheric Imaging Assembly in the Solar Dynamics Observatory (SDO/AIA) requires the availability of computational methods able to detect, trace and analyze numerous phenomena like flares, filaments, coronal mass ejections and active regions. OCCULT-2, an automated pattern recognition code for the extraction of one-dimensional curvilinear features from two-dimensional images, has proved a notable effectiveness in the identification of loop structures in SDO/AIA data. We now present a novel approach for the automatic determination of the mathematical equations describing these structures that relies on the results provided by OCCULT-2. Specifically, our approach implements a generalization of the Hough transform procedure to the recognition of patterns described by algebraic curves. In this method the points determined by OCCULT-2 are transformed into either curves or surfaces in the parameter space and an accumulator procedure is used to optimize the parameter values that recognize the curve in the image space. We validate this procedure against synthetic but realistically simulated data and show its effectiveness in the analysis of a number of SDO/AIA images.