Abstract: The morphological characteristics of aurora are strongly affected by the contemporaneous solar wind variance and interplanetary magnetic field. Large amounts of information about the magnetosphere and solar wind activities can be got through the study of aurora morphology. With millions of aurora images collected, how to classify the aurora sequences fast and efficiently becomes an important and hot topic. At present, the dynamic features of aurora sequence are seldom extracted compare to the static features of aurora images. In this paper, a method for classifying aurora sequences based on the dynamic texture model is proposed. Linear dynamical texture systems are used to model all four categories of aurora sequences universally. The scheme can acquire the dynamics and correlations between aurora frames. Then SVD decomposition of matrix was used to the solution of dynamic texture model. Finally, the difference between sequences are evaluated by the Martin distance of model parameters. The nearest neighbor(NN) classifier and SVM classifier is applied to classify aurora sequences. In order to reduce the redundancy of the model, tensor dynamic texture models are used based on tucker tensor decomposition. Compare to dynamical texture models which only focus on the correlations between the aurora frames, tensor dynamic texture models also analysis the correlations in one frame, because tensor dynamic texture models decompose the aurora sequences both on spatial and temporal dimensions. Then the parameters are extracted as the features of aurora sequences, which reduce the redundant information while enhance the classification accuracy at the same times. The experimental results on the dataset obtained from All-sky Imager (ASI) at the Chinese Yellow River Station achieve high classification accuracy, and demonstrate the effectiveness of the proposed classification algorithm.