Optical Flow-Based Dayside Aurora Image Classification

The auroral phenomena is formed by solar wind collides with the atoms in the upper atmosphere over the earth. The research of spatial structure and temporal evolution of the auroral can obtain many information of the solar wind-magnetosphere interaction and the physics of the magnetosphere-ionosphere interaction. Dayside auroral is the ionosphere track generated by the solar wind-magnetosphere interaction, reflects the interaction of solar wind and magnetosphere in dayside. In addition it is of great importance to the study of space weather activity, ionosphere and its dynamic features.In this paper,we analysis the latest achievement in auroral classification. We try to class the auroral images and aurora image sequences by their motion information.According to the non-rigid feature of dayside corona aurora, we proposed optical flow for the fluid motion to compute the auroral image motion, then introduce a multi-resolution scheme in optical flow to resort the rapidly change of auroral image.According to the large data of optical flow information, we introduce Local Binary Pattern algorithm base on optical flow filed partition. Then k-NN classifiers are employed to realize the classification of aurora. Finally we designed the experiments on the auroral data from Huanghe station, the results illustrate the effectiveness and feasibility of the proposed algorithm.Based on the above we proposed a detection method of aurora image sequences based on the optical flow field feature. Because the image sequences with time information, we connect the Local Binary Pattern feature of optical flow field in to describe the image sequence, the experimental results illustrate the effectiveness and feasibility of the proposed algorithm.