The invention discloses a video sequence classification method based on a viscous fluid particle motion model, which is used to solve a problem of inability of classifying an aurora sequence by using motion characteristics. The video sequence classification method comprises steps that (1) the input aurora sequences are preprocessed; (2) motion fields of preprocessed aurora particles are calculated by using the viscous fluid particle motion model; (3) the local binary pattern characteristics of the aurora particles are extracted, and are used as the dynamic characteristics P1 of the aurora sequences; (4) the local binary pattern characteristics of the pixel value of every frame aurora image of the aurora sequences are extracted, and are used as the static characteristics P2 of the aurora sequences; (5) by combining P1 and P2, the characteristics P+(P1,P2) used to express the aurora sequences having different forms are acquired, and the characteristics P of the aurora sequences having the different forms are input in a support vector machine to complete classification. The video sequence classification method is advantageous in that the automatic aurora sequences is realized; classification accuracy is high, and classification speed is fast; the video sequence classification method is used for scene classification and target identification.