Clustering Video Sequences by the Method of Harmonic k -Means

This study is devoted to the segmentation–clustering of video sequences by analyzing multidimensional time sequences. An approach is proposed to using the iterative deepening time warping technique in conjunction with the matrix harmonic k-means method. Unlike the traditional approach, this segmentation–clustering procedure is insensitive to the initial choice of centroids, which is especially convenient when analyzing arbitrary mass data.