Automatic Detection Of Aurora Based On LBP-SVM

The aurora phenomena provide a convenient projection for the study on complexenergetic plasma processes of the outer magnetosphere. The spatial structure andtemporal evolution of auroral luminosity are ascribed to cumulative effects of the solarwind–magnetosphere interaction and the physics of the magnetosphere–ionosphereinteraction. Therefore, the study on auroral appearance and evolvement is helpful to findout influence of the sun to the earth, and is significant to acquire information aboutspace weather.It has been paid more and more attentions that pattern recognition technologies areused in aurora data processing, such as aurora detection, aurora classification, etc. Inthis paper, firstly, with the last achievements in aurora detection, an automatic detectionalgorithm base on auroral arc model has been reviewed. Secondly, for the difference ofthe image brightness, an algorithm based on the brightness threshold has been proposed.Lastly, we use the local binary pattern (LBP) to extract the features of the auroraimages, and use support vector machine (SVM) to complete aurora recognition. Theexperiments are carried on the auroral data from Huanghe station, and the resultsindicate the effectiveness and feasibility of the proposed algorithms. Compared to thoseon auroral arc model and the brightness threshold, the algorithm based on LBP-SVMhas a higher detection rate.